Explore with evidence

Find roads worth exploring.

Choose a place, find the roads nearby worth a closer look, and see the mapped natural features behind each one.

01 / Discovery

Discover roads

0

02 / Geography

Corridor map

The map shows the same filtered and sorted corridors as the candidate list, over OpenStreetMap tiles with their attribution. Geometry is evidence, not an invented route, and if the tiles cannot be loaded the corridors are still drawn over a blank background.

03 / Candidates

Road corridors

0

06 / Saved

Saved roads

0

07 / Outings

Outings

0

08 / Field trips

Recorded trips

0

04 / Investigator

Evidence & questions

About this workspace

What this is. Road Naturalist is a road-exploration workspace over mapped natural features. Find roads worth exploring from mapped water, wetlands, ecology and road context; see them together on the map; understand why each one surfaced; and keep the promising ones for an outing. Discovery measures facts about a corridor and offers one documented exploration order built from that physical context — the five recommendations a search answers with are the first five roads in it. It does not know whether wildlife is present, whether a road is scenic, whether it is legal or safe to drive, or whether it is objectively better than another road. The demo pilot area uses real road, habitat and occurrence data, and its access stays unverified.

Road Naturalist is an early research workspace. The pilot corridors use real road-centerline geometry from the U.S. Census Bureau TIGER/Line 2025 ROADS dataset, real U.S. Fish & Wildlife Service National Wetlands Inventory polygons, and real U.S. Geological Survey hydrography lines, so their EPA Level III/IV ecological context and their wetland and surface-water surroundings are calculated from real Oregon GIS data. Nothing here is a recommendation, a wildlife record, or a habitat-quality score; the access findings are dated, sourced, qualified evidence rather than a legal determination.

Place names are a lookup, not a geocoder. Typing a name in discovery searches a small local list of named places in the published region and a 50-mile margin around it, reduced offline from the U.S. Census Bureau Gazetteer file: incorporated cities and towns, and census-designated places, with the Census Bureau’s published interior point as the centre. It holds no street addresses, landmarks, streams or peaks, it never contacts a geocoding service or your location, and choosing a place only sets the search centre — the radius stays yours to choose, and a shared link always carries coordinates.

Geometry is not access. A named road in a government centerline dataset proves that the road is mapped, not that the public may legally or practically drive it. A mapped road/water intersection is a geometric crossing, not a bridge, a ford, or current water. Access stays unverified until the Investigator checks it. Species occurrence evidence comes from public iNaturalist and eBird reports: it records what people publicized, at the precision the source publishes. Obscured or imprecise observations stay regional evidence and are never measured against the road or plotted, and a public report is not a prediction that a species is on the road. Occurrence sources are queried only when you ask; eBird needs a server-side credential that this browser build does not hold, so it reports UNKNOWN rather than zero.

Access is a qualified finding, not a switch. The Investigator researches whether a road is a public road with ordinary motor-vehicle access and reports one of: verified public, probable public, unverified (the default), conflicted, or restricted/closed. It preserves the restriction itself (temporary, seasonal, permit-only, gated) and it never resolves a contradiction between sources by preference. Access evidence carries its source class (an authority acting in its own jurisdiction, government data, community mapping, or anecdotal web evidence), the verbatim text it came from, its retrieval date, and its effective dates, because a closure notice ages differently from a road centerline. OpenStreetMap is supporting evidence: an absent access tag means the tag is absent, not that the road is public. Every finding shows the deterministic guardrail rule that produced it, the adversarial concerns raised against it, and what remains unresolved — and a human finding is recorded beside the automated one rather than replacing it.

Discovery offers one documented recommendation order, and it is not a prediction. Candidate discovery reads a bounded road-network extract, keeps only the road classes TIGER/Line classifies as ordinary vehicular roads, joins source features into named roads, divides long roads into contiguous analysis corridors, and then measures road, wetland, surface-water, and ecoregion facts for every corridor in one set-oriented GIS pass. The five recommendations a search answers with are the first five roads in the exploration order documented in docs/DISCOVERY.md: mapped wetland, water and ecological variety, with a small, bounded preference for roads nearer the search centre. It shows the measured values and their coverage so they can be filtered and sorted; it does not predict wildlife. It never queries species-occurrence APIs or official access sources — those begin only when a person promotes a corridor for deeper investigation, and a promoted corridor needs no reviewed research source to exist: without one, its access stays UNVERIFIED and the Investigator reports that no reviewed sources are declared.

Future GIS results will carry source, version, coverage, and computation details. An Investigator may interpret and check them, but cannot create authoritative facts.