@rhhs “cohesive dust” fixed to “fine dust”. Will add clay to the Key:surface unpaved section as a separate edit, and a firmness reference once that proposal progresses.
@9_tab By that logic surface=grass should differ between a soccer pitch and a meadow path The “Note on terminology” handles this: feature type (leisure=pitch vs highway=*) already provides the context.
@ftrebiensurface=ground dropped from both the fallback and the “when uncertain” sentence, surface=dirt only. The other values you mentioned (grass, sand, gravel, mud) aren’t realistic confusion cases with clay, if you can’t tell whether it’s clay, it’s dirt. On the opening paragraph: used “difficult or impassable” rather than “impassable” since passability varies by conditions and vehicle.
There’s also artificial clay which is a plastic carpet filled with sand, on top of an asphalt surface (video). There are a lot of them, but surface=artificial_clay has only been used 17 times, so it is likely these have generally been tagged surface=clay too.
I hope we can agree that exposed clay soil (which you want to tag) and “clay” tennis courts look completely different and behave completely differently underfoot. The surface I am talking about is not a subtype of ground or dirt. It’s closer to compacted.
In principle I agree with you that if people consistently used surface=clay in one meaning on sports courts and in the other meaning on highways, then there isn’t a high risk of conflict, because then a data consumer can determine from the context (tennis court or road?) which meaning the mapper had in mind. But shouldn’t we still try to avoid actively creating situations where the meaning of a tag changes depending on which other tags are present?
And also I am not so sure that it’s so clear cut. We need to look at how people have actually used surface=clay on highways. Here’s the distribution:
Only about 600 tracks, and very few “real” roads. But also 600 footways and 300 paths. Let’s look at footways. I didn’t have to look at many to find examples where people have used surface=clay to mean a tennis-court like surface.
Where people have used surface=clay on footways, they have at least sometimes done this because of its similarity with the tennis court surface. Examples are here and here. These are real examples where people have used surface=clay. They don’t show exposed soil. They show well compacted walkways covered by a thin layer of dust/sand. Personally I would have probably used a different tag but that’s what the mapper used. They were probably familiar with “clay” tennis courts and recognised this as a similar surface.
In short I don’t agree with your claim that you are just documenting actual usage. I think we have a tag that most of the time (> 90%) means something quite specific, and in a few, relatively rare cases has been used to mean something quite different. Your proposed page is mostly about that rare other usage, which you are trying to promote.
It’s a bit of an unfortunate situation because it shows geographic bias in where OSM started originally..
Proposals are optional, you can use any tag you like..
You could just start using surface=clay_soil or similar. Another idea might be surface=dirt dirt=clay, this would mean that data consumers will have a fallback if they haven’t implemented the new tag specifically. From a routing perspective, that’s maybe even better than using surface=clay as I don’t think that many routers support =clay. (From a quick search on Github, out of the three on osm.org, only Valhalla does.)
I wouldn’t be able to tell that this is clay without interacting with it. It looks like asphalt or concrete, but the fact that the player is sliding on it strongly suggests it is clay indeed.
Clearly not clay, surface=artificial_clay makes more sense. I guess this would not be used to make roads, but in any case, the narrator says it is “quite slippery compared to real clay”, indicating a likely important practical difference between the two surface types.
The usage split you found is real, and I don’t dispute the numbers. But I think the interpretation deserves caution.
Documentation gap
The ~2,000 highway uses exist under documentation that says almost nothing about roads. The current wiki page consists of two short paragraphs focused on tennis courts.
That’s not strong evidence that highway use is marginal, it suggests that mappers in clay-soil regions have been using the tag without guidance or presets. I personally know of many ways I haven’t tagged yet because I was waiting for clearer documentation. The numbers likely reflect documentation neglect more than mapper consensus.
The breakdown also matters. Ohsome data shows highway=residential and highway=unclassified, not just track. These are community roads in developing countries, tagged by local mappers describing their actual streets, not by people thinking of tennis courts.
This may be a semantic disagreement
“Clay” is first and foremost a geological material. In practice, soils are mixtures, and what we call “clay soil” is soil where clay is the dominant component controlling behaviour.
A road whose surface layer consists of clay-rich soil, whether lateritic clay in West Africa or red clay in Southeast Asia, is materially a clay surface in the ordinary sense of the word.
By contrast, a “clay court” in tennis is a constructed sports surface, usually made from crushed brick. That usage is established, but it is a sporting convention.
OSM surface tags describe the dominant material present (gravel, sand, grass), not chemical purity. Applying the same principle, a road dominated by clay-rich soil fits naturally under surface=clay.
So the key question is not which community uses the tag more often, but whether describing a clay-soil road as surface=clay is materially accurate. In those cases, it is.
Those examples are genuinely ambiguous. Many would likely be better described as surface=compacted, and the wiki can state that explicitly.
But ambiguous footways don’t invalidate the clearer cases, clay-rich roads whose defining behaviour, hard when dry, plastic and sticky when wet, prone to rutting, is precisely what surface=* is meant to capture.
Other material tags already span feature types:
surface=(fine_)gravel on pétanque courts and roads
surface=grass on soccer pitches and rural tracks
surface=sand on beach sports and sandy roads
OSM generally tags the material and lets context disambiguate.
surface=dirt + dirt=clay could be a refinement, but it shouldn’t be required for someone who can clearly identify clay-rich soil in the field. surface=clay already exists, is referenced from surface=ground, and has established highway use. Introducing clay_soil would fragment usage rather than clarify it.
The draft acknowledges both contexts and recommends surface=dirt when identification is uncertain. It does not promote ambiguous tagging, it clarifies when clay is appropriate.
If there are specific wording changes that would improve clarity, I’m happy to incorporate them. Otherwise, I plan to proceed with the wiki update.
To keep things balanced, I’m also willing to improve the general surface=* documentation to explicitly reflect that certain material values, like grass, (fine_)gravel, and sand, apply across both sports and transport features.
The goal is clearer documentation and consistency, not promotion of a niche reinterpretation.
To be precise, “gravel”, “sand” and “clay” can be about the same chemical “material”, like “pebble” and “silt”, just in different grain sizes. They are not detailing the chemical composition (they are all some kind or mixture of “minerals”). “dirt” is a very unspecific word compared to the aforementioned, and does not fit into the same classification.
Compared to the multiple uses of Key:emergency - OpenStreetMap Wiki I think the different meanings of surface=clay is manageable. Maybe a sentence can be added that “If the surface of a highway looks like that of a clay tennis court, use surface=compacted.” But only in the English wiki: looking at the versions in the languages I know of Clay court - Wikipedia , only English uses the word “clay” to describe that surface (German: sand; Dutch, Danish, Norwegian: gravel; French, Italian: beaten soil; Bulgarian: red).
If the concern is that surface=* is being used across very different feature types, then the structural solution would have been to reserve surface=* strictly for transport features and introduce something like material=* for everything else.
What seems less consistent is effectively duplicating material values inside surface=* based on context, or treating the same value as valid in one domain but problematic in another.
If surface=* is intended to describe the physical material present, then that principle should apply uniformly, whether the feature is a road, a pitch, or something else. Otherwise we end up with parallel semantics for the same key, depending on feature type, which makes the tagging model harder to reason about.
If usage of surface=clay on roads grows significantly over time, we could always revisit the idea of splitting the tag into more specific values through a formal proposal. For now, documenting both existing uses seemed like the most practical step.
I removed the description of the colour, because I think it’s not a very decisive indicator (it can be almost any soil colour), and added an image of the drying pattern.
The “Note on terminology” handles this: feature type (
