# Data Sources

Verified source links are listed in `docs/source_references.md`.

## Core principle

The raw route network can be built from OpenStreetMap. The competitive advantage comes from derived intelligence:

- route segment suitability,
- loaded-bike rideability,
- route-corridor logistics,
- water/food/sleep reliability,
- local rules confidence,
- trusted local reports,
- actual ride feedback.

## Data-source matrix

| Domain | Source candidates | MVP use | Production use | Caveats |
|---|---|---|---|---|
| Base map/network | OpenStreetMap extracts via Geofabrik or planet.osm | Road/path graph, surfaces, access, POIs | Full regional/global ingestion | ODbL obligations; data quality varies |
| Routing engine | openrouteservice, GraphHopper, Valhalla, BRouter | Hosted API or local engine adapter | Self-hosted customized profiles | Requires tuning; public APIs have limits |
| Elevation | Copernicus DEM GLO-30/GLO-90, NASA SRTM | Elevation profile and grade warnings | Segment-level grade/hike-a-bike modeling | DEM artifacts; bridges/tunnels need handling |
| Vector maps | OpenMapTiles, Protomaps/PMTiles, MapTiler, custom tiles | Web/mobile map display | Offline vector packs | Attribution and tile generation costs |
| POIs | OSM POIs, Overture Places, partner APIs | Water/food/sleep/repair corridor | Reliability scoring and deduplication | POIs stale/incomplete; licensing varies |
| Protected areas | Protected Planet WDPA/WDPCA, national/regional open data | Rule warnings | Rule confidence engine | Rules are jurisdiction-specific |
| Weather | Open-Meteo, MET Norway/Yr, national weather services, commercial APIs | Forecast along route | Weather-aware rerouting | Forecast uncertainty; API licensing |
| Accommodation | OSM campsites/shelters, Booking Demand API, direct campsite/hut partners, Warmshowers-style partnerships | Deep links and search cards | Reservation/check availability where allowed | API access/partner terms |
| Bike devices | GPX/TCX/FIT export, Garmin Courses API, Wahoo Cloud API | GPX export | Direct sync | API approval required |

## OSM tags to prioritize

### Route network

- `highway=*`
- `bicycle=*`
- `access=*`
- `surface=*`
- `smoothness=*`
- `tracktype=*`
- `sac_scale=*`
- `mtb:scale=*`
- `incline=*`
- `ford=*`
- `bridge=*`
- `tunnel=*`
- `barrier=*`
- `route=bicycle`
- `network=*`

### Water

- `amenity=drinking_water`
- `drinking_water=yes/no`
- `natural=spring`
- `man_made=water_well`
- `amenity=fountain`

### Food and resupply

- `shop=supermarket`
- `shop=convenience`
- `shop=bakery`
- `amenity=cafe`
- `amenity=restaurant`
- `amenity=fast_food`
- `amenity=fuel`

### Sleep

- `tourism=camp_site`
- `tourism=caravan_site`
- `tourism=alpine_hut`
- `tourism=wilderness_hut`
- `tourism=hostel`
- `tourism=hotel`
- `amenity=shelter`

### Bike repair

- `shop=bicycle`
- `amenity=bicycle_repair_station`
- `service:bicycle:repair=yes`
- `service:bicycle:pump=yes`

### Bailout

- `railway=station`
- `public_transport=station`
- `amenity=bus_station`
- `route=ferry`

## Data ingestion notes

### OSM

Use regional PBF extracts for development. Do not query public Overpass for production-scale route-corridor POI needs. Import regional extracts into PostGIS or a search index.

### POI deduplication

POIs may appear in multiple sources. Use a deduplication key based on:

- geohash proximity,
- normalized name,
- category,
- phone/website if available,
- source reliability.

### Reliability scoring

Each POI should have:

- `source`,
- `last_seen_at`,
- `last_confirmed_by_user_at`,
- `seasonality`,
- `opening_hours`,
- `confidence`,
- `bikepacker_notes_count`,
- `negative_reports_count`.

### Rules

Rules must be jurisdiction-specific and confidence-scored. Never present legal/camping rules as absolute unless sourced from official/local authority data and current.