howsteep.org — reproducible LiDAR gradient ladders for road climbs, so far in the UK, the north-eastern United States, Hawaii, Switzerland and Austria: wherever open 1 m (or finer) LiDAR and an open road inventory exist, the survey can go. Austria is measured without downloading a terrain model at all: the federal 1 m grid is read block by block from its server, one kilometre at a time. Every maximum states its measurement distance, because a hill can honestly be 32% over 2 m and 25% over 100 m at once.
This repository is the public face of the project: corrections, local knowledge and photo reports are collected here as issues. The measurement pipeline lives in a private repository.
Use the climb report form. The most valuable reports are the ones only a person on the ground can make:
- a gradient sign we don't show, or one that has been repainted (signs change — we date every observation);
- a road that is gated, private or unsurfaced despite our checks;
- a name locals actually use for a climb we auto-named;
- a climb we're missing entirely;
- photos of any of the above — drag them straight into the form.
Discovered climbs on the site are hypotheses derived from LiDAR terrain data and open road datasets. Despite extensive computational verification, some can still be artifacts of the terrain model or the map — that's exactly what your reports help fix.
Why isn't my hill ranked? Every road the survey has ever considered is in the decisions dataset: published and ranked, published but held for a person's look, withdrawn, merged, measured but never given a page, or proposed and never measured — each with a coded reason and the rule behind it. Look your road up there first; if the rule got it wrong, that is exactly what the report form is for. The codes are explained in DECISIONS.md.
The site's measurements are published in OpenClimb Ladder (OCL) — a data format in which a maximum gradient cannot be expressed without its measurement distance. The schema forbids windowless maxima; that is the point. Spec: spec/OCL.md · normative JSON Schema: schema/ocl-climb.schema.json · live example: any climb page on howsteep.org links its own OCL document. Comments and implementations welcome — open an issue.
Format changelog (all changes within 0.1.x are additive):
- 2026-09-07 — spec and schema brought level with the live corpus:
provenance.run_id(opaque batch-run identity, spec section added 2026-08-22); per-window lateral-stability fields on ladder entries —placement_sensitivity_pp,estimated_best_pct,estimated_range_pct,estimate_basis— recording how much a maximum moves when the measurement line is shifted across the carriageway and the resulting estimate of the true reading. Documents published since 30 August carry them.
Measurements are derived from open data: national LiDAR programmes of all four UK nations (OGL v3), OS Terrain 50 / Open Names / Open Roads (OGL v3), USGS 3DEP LiDAR (public domain) with the states' road inventories, and OpenStreetMap (ODbL). What the survey publishes, you may reuse — the terms are settled:
- any individual climb's page, charts and figures — CC BY 4.0, with credit to "HowSteep — howsteep.org";
- the complete dataset (CSV and per-climb OCL documents, downloadable from howsteep.org) — ODbL 1.0 (attribution + share-alike, as inherited from OpenStreetMap);
- the decisions dataset (every route ever considered, with its reason code; tables and per-state geometry at howsteep.org/data/decisions, the full GeoPackage on this repository's Releases page) — ODbL 1.0; see DECISIONS.md;
- the OCL spec — CC BY 4.0; the JSON Schema — MIT;
- the site as a whole (its selection, arrangement and complete corpus) — not licensed; for bulk use beyond the ODbL dataset, get in touch (a private message to the editor; no account needed).
Details and attribution guidance: howsteep.org/licensing.html · LICENSE.md.