Working version, 9 October 2026
The tool finds every place in Agder where a road or path crosses a mapped stream, river or canal. It checks which of those places have a culvert or bridge in the official road database, and treats the rest as assumed culverts.
Everything comes from public sources. The tool does not need any field data to run. Field data is only used to check the result.
| Data | Owner | What it gives the tool | Size in Agder |
|---|---|---|---|
| Roads and paths | |||
| NVDB road network | Statens vegvesen | All road types: European (E), national (R), county (F), municipal (K), private (P) and forest roads (S), plus foot and cycle paths | 170,000 segments, 17,114 km of road |
| OpenStreetMap | Volunteers (Geofabrik extract, 6 Oct 2026) | Roads, tracks and paths that NVDB does not have | 108,600 ways, +2,564 km of road, 5,894 km of paths |
| Water | |||
| FKB-Vann streams and canals | Kartverket | The main stream map: precise centre lines, with a height on every point | 772,940 lines |
| FKB-Vann wide rivers | Kartverket | Rivers wide enough to be drawn as areas | 20,945 areas |
| NVE Elvenett | NVE | A coarser stream network, used to fill places where FKB-Vann has no stream | 134,193 segments, 32,316 km |
| NVE lakes | NVE | Keeps stream lines that run through lakes from counting as crossings | about 23,000 lakes |
| Registered structures | |||
| NVDB bridges and culverts | Statens vegvesen | Which crossings are registered, with diameter, length, material and whether a stream runs through | 3,507 bridges, 54,353 culverts |
| Used for checking only | |||
| Field survey of culverts | Agder field survey | 718 surveyed culverts and bridges, each rated for fish passage | 718 structures |
| NINA terrain study (report 2519) | NINA | An independent method based on laser-scanned terrain, used as a second opinion in two test areas | 2 test areas, 7 × 7 km |
Seven steps run in order. Each step lists the settings it uses.
Every point where an NVDB road line meets an FKB-Vann stream or canal line becomes a crossing. Several hits close together are one real crossing.
The same test with OpenStreetMap roads, tracks and paths. A crossing is dropped if the first step already found one close by. State and county roads are left out of this step, since NVDB covers them well.
Stretches of NVE Elvenett that lie far from any FKB-Vann line are tested against the roads. Roads crossing the wide-river areas are added too. Lakes are masked out first.
FKB-Vann often ends the stream line at the road where the water goes into a pipe. If a line ends on one side of a road and another line starts on the other side, and the water would flow from the first to the second, it counts as a crossing. Flow direction comes from the heights stored on the lines.
Each crossing looks for a registered bridge or culvert nearby, in NVDB or tagged in OpenStreetMap. Found and it is a culvert: registered culvert. Found and it is a bridge: small or large by length and span. Not found: assumed culvert. A road crossing a wide river area with nothing registered is treated as a probable bridge.
Road type, with the highest class winning if several meet: European, national, county, municipal, private, forest road, road only in OpenStreetMap, path. Water type: stream or canal and ditch. Municipality.
NVDB has 54,000 culverts. Many sit where the water maps show no stream, mostly road drainage. They are listed on their own, switched off in the map by default.
Most values were chosen from how the data behaves. The 30 m radius and the large-bridge limit were checked against the field survey.
| Setting | Value | Reason |
|---|---|---|
| Registered-culvert search radius | 30 m | Of the surveyed culverts that lie within 50 m of one of our crossings, 98% (653 of 666) are already within 30 m. A wider radius adds little and risks matching the wrong culvert. |
| Large bridge | 15 m long or span | 27 of 91 surveyed registered bridges were rated partial or absolute barriers, almost all shorter than 9 m. None of the six bridges of 15 m or more was. Short registered bridges therefore stay visible. |
| Merge hits | 15 m | A road and a ditch beside it can hit the same stream a few metres apart. |
| Same crossing from two sources | 25 m | Positions in NVDB, OpenStreetMap and the water maps differ by some metres, so one crossing can show up twice. |
| Elvenett gap | 50 m | Elvenett is a coarser map, so closer lines are treated as the same stream already in FKB-Vann. |
| Stream-end link | 45 m, 65°, 1.5 m | Wide enough to span a road with verges, tight enough to stop two unrelated streams being joined. The height rule keeps water from flowing uphill. |
Share of crossings by status, for each road type.
| Road type | Crossings | Share by status | Assumed | Road length |
|---|
On national and county roads almost all crossings are registered (only 4% and 6% are assumed). On municipal, private and forest roads and on paths, 69 to 89% are assumed. That matches how the data is kept: owners of those roads are not required to register culverts.
Of 718 surveyed culverts and bridges, 91% lie within 30 m of one of our crossings. The share is the same for culverts rated as barriers and those rated no barrier.
Limit: 96% of these are on county roads. The survey says little about municipal and private roads, where the largest gap is.
In two 7 × 7 km test areas, 67–82% of our crossings have a terrain-derived culvert point within 25 m (84–95% within 50 m).
The terrain method finds 8 to 11 times as many points, mostly small drainage lines. It cannot say which of them carry a real stream.
Manual checking of the map so far has found few mistakes, mostly on small waterways. A filter separates streams from canals and ditches, which are often roadside ditches.
The first filter planned is reachability for anadromous fish, using Miljødirektoratet's open datasets on anadromous reaches and migration barriers. Each crossing will also need to carry the nearest stream in the NVE network, so the route down to the sea can be followed.
Sources: Statens vegvesen (NVDB), Kartverket (FKB-Vann, municipal boundaries, topographic map), NVE (Elvenett, lakes), OpenStreetMap contributors (via Geofabrik), NINA report 2519 (2025), field survey of culverts in Agder. All counts are for Agder (fylke 42), from the 9 October 2026 run.