The MTB classification on the OpenStreetMap map (mtb:scale and the tags that go with it) is a genuinely usable indicator of how rideable a trail is for mountain biking. The classification makes route planning a great deal easier: the mere existence of a classification already tells you that some mountain biker has ridden the trail and seen fit to classify it. And the classification itself tells you, with reasonable accuracy, how rideable the trail is.

There has been the occasional debate about how accurate the MTB classification is and how hard objective assessment is: a yellow trail in one area and terrain type (an esker area, say) may be considerably easier than in another (a classic mixed forest, more rocks and roots). Differences and margin of error aside, the MTB classification is in practice a reliable enough basis for route planning: if you want something really easy, you pick green and a bit of yellow and steer clear of orange --> there is a good chance a beginner will enjoy the route and will be riding the bike rather than pushing it.

The gravel route problem

For gravel riding, things have not been in nearly as good shape. There are of course usable ways easy to find on the map for gravel (bigger gravel roads, gravel-surfaced recreational paths, easy trails) — but making full use of gravel's advantages, "ride almost anywhere, the most varied cycling routes there are", is harder. And the more adventurous the routes you plan, the more easily you end up on ways where riding a gravel bike is either very slow and hard work, or outright impossible.

The gravel route problem could be summed up like this: on big roads (the ones cars use) and in urban areas the character of the ways is clear, and at the other extreme, on MTB-classified trails, there is at least a reasonable guess at rideability (even though MTB and gravel differ significantly when it comes to riding trails). But in between, forest tracks, ski trail beds and some recreational paths are trickier, because their quality is not as easy to deduce from the map — and when you use automatic routing, the problem is bigger still for the routing engine.

Not every way you find in the forest is champagne gravel...

Planning a gravel route

The big opportunity in gravel route planning is "a new route for every ride", offering the chance to comb through the ways thoroughly and to ride even familiar ways in ever new combinations. Making that possible calls for the ability to plan routes fast: a route thrown together with a few clicks before the ride, and an app interface that tells you quickly and clearly whether the route has spots you need to check and possibly reroute around.

The newest version of Trailmap answers that need by still offering the familiar, fairly bold gravel routing — routing that is not too shy about using even the smaller ways, but that also knows how to sensibly avoid, say, trails that are too awkward. The new version shows a new "gravel classification" for the ways, and the warnings that go with it. The warnings show up directly as colours in the elevation profile, and clicking the elevation profile takes you straight to the clicked spot on the route. That way the spots worth checking are visible at a glance and checking them is quick and easy.

The gravel classification

Trailmap's new gravel classification is "virtual", i.e. it is not based on any single new OSM tag (there is no gravel:scale in OSM data). Instead, the existing OSM data is interpreted as accurately as possible and the gravel classes are derived from that:

Class	Description	OSM tags
0-	Asphalt	surface & highway
0	Cycleway (gravel) or presumed (but not certain) asphalt	highway & tracktype=grade1
0+	Gravel roads, recreational paths and the like (risk of grading!)	numerous different OSM tags
1	Often equivalent to class 0+, but a notch more risk	numerous different OSM tags
2	Ok as short connectors, but slow / risky for longer stretches	numerous different OSM tags
3	Unknown, and the risk cannot be assessed	e.g. highway=track with no other tags
4	Identifiably very demanding or unrideable	numerous different OSM tags

The guiding ideas behind the classification:

• Gravel is "all-terrain" cycling, i.e. asphalt and trails fall under the definition too
• With that definition, sub-types can be derived from the gravel classification: road cycling, touring (big fast ways), easy gravel, technical gravel and so on
• For gravel riding in general the aim is to enable what you might call managed risk-taking: the route can have more demanding stretches, but only so that the planner knows about them and can think about, say, a "plan B" if some forest track turns out to be too awkward
• On gravel ways the conditions change: maintenance (grading / regravelling, or on forest tracks reinforcement with coarser "gravel") and the season (wet weather in autumn / spring, and sucking gravel) — and OSM data cannot account for the effect of that. That is why the gravel classification has classes 0, 0+ and 1 — which in terms of rideability may often be exactly the same, but whose risks nonetheless clearly differ

The OSM tags used are: highway, surface, mtb:scale, tracktype, smoothness, obstacle, access, bicycle, width, piste:type, trail_visibility. Combinations of these tags are used to try to identify the properties of a way. Since there is no single "mtb:scale"-type tag available, and given the variable condition of gravel ways, deducing the classification from the tags is an exercise in estimating probabilities.