Source fact
A field or a geometry published by an identified public authority, stored as the authority published it. We keep every column a source exposes, and we keep the date it published. If we did not receive it, the page does not carry it.
Data & methodology
Every field has a source and a date.
Deep Oʻahu republishes public records. This page is the inventory of what those records are, the model that keeps a published fact separate from a calculated one, and the path for correcting anything that looks wrong.
7public-record sources
17datasets ingested from them
~320public-record fields captured
$0spent on data licensing
Sources
What the site is built from.
Every dataset below is free public record, pulled from the authority that publishes it. Nothing here is licensed, purchased, scraped from a competitor, or contributed by a user.
| What it answers | Publisher | Dataset | Records | Cadence | Last pulled |
|---|---|---|---|---|---|
| Parcel boundaries and TMK | State of Hawaiʻi, Office of Planning | ParcelsZoning — Oʻahu Parcels | ≈172,000 parcels | Weekly rebuild | Published on each page |
| Ownership, tenure, assessment, building, lot | City & County of Honolulu, Real Property Assessment Division | CadastralTables — OWNDAT, OWNERTYPE, ASMTPITT, LEGDAT, FACILITIES | ≈326,000 properties | Weekly rebuild | Published on each page |
| Zoning class and its description | State of Hawaiʻi, Office of Planning | ParcelsZoning — County Zoning | 1,965 county zoning polygons | Weekly rebuild | Published on each page |
| Flood zone and Special Flood Hazard Area status | FEMA, republished by the City & County of Honolulu | HNL_Flood_Zones | 1,714 FEMA flood-zone polygons | Weekly rebuild | Published on each page |
| Sea-level-rise exposure | State of Hawaiʻi | SLRXA — 3.2 ft inundation overlay | Coastal inundation mask | Weekly rebuild | Published on each page |
| Cesspool evidence and replacement priority | University of Hawaiʻi Sea Grant; Hawaiʻi Department of Health | OSDS points; cesspool priority bands | 10,805 mapped cesspools on Oʻahu | Weekly rebuild | Published on each page |
| Registered short-term rentals | City & County of Honolulu | STR_Tax_Point; STR_Units | ~1,400 registered short-term rentals | Weekly rebuild | Published on each page |
| Permit history | City & County of Honolulu, Department of Planning and Permitting | data.honolulu.gov — Building Permits | 432,021 building-permit records, 2005–2025 | Weekly rebuild | Published on each page |
| Sales records | City & County of Honolulu, Real Property Assessment Division | County deed-transfer record | Recorded transfers per property | Weekly rebuild | Published on each page |
| Walkability | US Environmental Protection Agency | National Walk Index | Index per census block group | Weekly rebuild | Published on each page |
| Traffic volume near a property | FHWA Travel Monitoring Analysis System; Hawaiʻi DOT | TMAS station counts; HPMS road segments | 51 Oʻahu counter stations; observations 2011–2025 | Weekly rebuild · partial road coverage | Published on each page |
| Reported traffic incidents | City & County of Honolulu | data.honolulu.gov — Traffic Incidents | Rolling dispatch record | Twice daily — not live | Published on each page |
| Aerial imagery under the map | City & County of Honolulu | 2024 orthophoto, self-hosted | 7.94 cm/pixel at the deepest level | One capture — 2024 | Published on each page |
On “last pulled”. The ingest ledger is a database table, and these pages are built ahead of a request, so stamping a date here would be stamping the date of the build rather than the date of the pull. Instead, every field on a property page carries the date its own source published it, next to the value. That is the date that decides whether a record is current, and it is the one worth reading.
Property model
A parcel is not a property.
Getting this wrong is why other tools show you a tower and describe it as a house.
Deep Oʻahu joins everything around the Hawaiʻi Tax Map Key. A parcel is a land lot with map geometry. A property is a separately owned, separately sellable thing — which may be the parcel itself, or a condominium/CPR unit that shares a parcel with hundreds of others.
So the island is ≈172,000 parcels and ≈326,000 properties, and those two numbers are supposed to disagree. Each property gets its own page, with its own owner, its own assessment and its own tenure, sitting on the parcel it actually occupies. No polygon is invented to give a unit a boundary it does not have.
- PullEach source is pulled whole. Every column a dataset exposes gets its own typed column — no opaque blobs, so a field can always be traced back to the row it came from.
- FuseRecords are matched by Tax Map Key and fanned out across CPR units, into one PostGIS store that is the only source of truth on the site.
- ServeThe search index and the map archives are rebuilt from that store. A search result or a map tile is a serving copy, never a second version of the truth.
Evidence model
Four kinds of evidence, kept apart.
A published record, a calculation over it, a model's observation and a picture are not the same claim, and a product that renders them identically is asking you to trust the weakest one.
Deterministic derivation
A reproducible calculation over source geometry: does this parcel intersect the inundation overlay, how far is the nearest counter station, which properties sit on this parcel. Same inputs, same answer, every time — and the calculation is named on the field.
AI-assisted observation
A model inference, such as classifying what is visible in imagery. It carries a confidence value and the evidence it was drawn from, and it is labelled an observation rather than a fact — never as a published record.
Generated visualization
A rendered illustration of something, rather than a source photograph of it. It is labelled as rendered wherever it appears, and it is never presented as evidence of a physical condition.
How that reads on the page
The sentence under a field says which class produced it. These four patterns are fixed — they are the same words everywhere on the site.
- 01
Source fact
Land tenureFee simpleCity and County of Honolulu, Real Property Assessment Division · 2026 assessment roll. - 02
Deterministic derivation
Flood zoneZone AE — Special Flood Hazard AreaCalculated from the mapped parcel boundary and the FEMA flood-zone polygon. - 03
AI-assisted observation
What is visible from the propertyOpen ocean sector, partially obstructedModel observation · confidence 0.7 · not a verified fact. - 04
Generated visualization
Illustrated contextRendered, not capturedRendered illustration, not a source photograph.
Source rights
Being able to reach data is not the same as being allowed to sell it.
Deep Oʻahu keeps a source-by-source record of the basis on which each dataset is used. A source enters the product when there is a documented basis for the intended use, and the basis is kept with the dataset rather than in somebody's memory.
The analytical layer is deliberately source-neutral, so a rights-cleared aerial, LiDAR or government input can feed the same calculation without the product being rebuilt around it.
The same rule runs in the other direction. Where a licence permits display but not redistribution — a utility's own planning tool, for instance — Deep Oʻahu links out to the authority's page rather than republishing the layer.
Nothing on this site is derived from a source that has not cleared that check.
Corrections
If a record here is wrong, tell us.
Public records change, and sources correct themselves. Deep Oʻahu rebuilds from stored authoritative data as updates arrive, so a correction at the source reaches the site on the next rebuild.
There are two different problems, and they have different fixes. If the county's own record is wrong, the correction has to be made with the county — we republish what they publish, and a change here would put the site out of step with the authority. If we have misread, mismatched or mis-dated a record that the source has right, that is ours, and we want to know.
Either way, send it. We will tell you which of the two it is, and where it goes.
- The pageThe URL you were looking at when you found it.
- The propertyThe address or the Tax Map Key, so we can find the same row.
- The fieldWhich value is wrong, and what it should say.
- The authorityWhere the correct value is published, if you know. A link to the county record settles it fastest.
Check it against a property you already know.
That is the only test of a data product that counts. Open the map, look up an address whose history you can verify, and see whether the record matches what you know.