I’m proposing surface=laterite as a documented value for the surface key on unpaved roads and tracks surfaced with lateritic soil, the red/orange tropical soil found across Africa, Southeast Asia, and Latin America.
Its key characteristic is a two-phase wet behaviour: on first rain the surface becomes near-frictionless within minutes, then under sustained wet turns deeply plastic and sticky, making it potentially impassable. Generic tags like surface=dirt don’t capture this, leaving routers with no actionable signal.
The proposal documents field identification criteria, distinguishes laterite from surface=clay and surface=compacted, and addresses a bot migration that silently removed hundreds of existing surface=laterite tags with no fallback. Community outreach reached 100 mappers across 16+ countries; 38 confirmed the field description or provided regional context.
Thanks for undertaking the effort to gather information from all over the world and pack it into this proposal. As you know I fully support the use of surface=laterite and I think it is high time that we take this material into account in our tagging scheme as it is the most common surface stuff in in the tropical world. Well done!
“Just use surface=dirt + material=laterite”: Tested in practice and failed: hundreds of highway transitions replaced surface=laterite with material=laterite, none added a fallback surface=* tag, and routers lost all signal.
this does not actually explain why surface=dirt + material=laterite is a bad tagging for this situation and why surface=laterite is better than that
(I am not claiming which one is better)
Rejecting this proposal does not produce cleaner tagging; it preserves the clay/laterite conflation.
why tagging surface=dirt + material=laterite preserves the clay/laterite conflation?
“This adds complexity for routers”: Routers already handle a wide range of unpaved surface values; this slots in at the same level as surface=dirt. The known wet-season behaviour gives routers more actionable signal than a generic dirt tag.
this is also not very clear why actually surface=laterite is superior
BTW, some basic plan to reduce negative impact on data consumers would be useful
may be as simple as creating issue at each project listed at surface=dirt | Tags | OpenStreetMap Taginfo - ideally at RFC stage to give chance to make feedback
Fair correction. The bot moved surface=laterite to material=laterite without adding any fallback surface=* tag, leaving the surface key absent, “removed without fallback” described the routing effect, not the literal operation. I’ve tightened that in the proposal and added that the bot is now stopped.
@Map_HeRo puts it well: no one argues for surface=paved + material=asphalt over surface=asphalt. The surface key exists for direct material values; the compound form is a workaround that hands consumers a generic signal regardless of what the mapper knew.
Routers query surface=* for penalties and don’t parse material=* in that context. The bot migration also showed the compound tag doesn’t propagate in practice: 711 transitions to material=laterite with zero fallback surface=* tags added.
On data consumer outreach: I’ve added a post-approval outreach plan to the proposal listing the projects to be notified: JOSM, iD, Vespucci, MapComplete, StreetComplete, Valhalla, OSM Carto, Shortbread, Americana, OsmAnd, Guru Maps. The proposal process sequences this after approval: consumer contact is recommended once there is a documented, community-approved tag for developers to act on.
Thanks for processing feedback! Reaction here look basically perfect (and I know how annoying/time-consuming it can get, I hope that my feedback is at least sometimes useful)
I will take look at that later, and comment if I will have something useful to say
surface=mud: a condition, not a material. Laterite alternates between firm and plastic seasonally; use surface=laterite regardless of current season.
I think that bigger difference is that surface=mud is applicable in places that have mud all the time (or nearly all the time, basically never drying). Making it actually material than a temporary condition. Maybe
surface=mud: it is for places that never or basically never are dry. Laterite alternates between firm and plastic seasonally; use surface=laterite regardless of current season.
would be better?
maybe also with
surface=mud would be applicable in rare cases where laterite is muddy entire year, not becoming dry at all
Do not infer laterite from colour alone outside tropical/subtropical context.
is it safe even in tropics? I expect that context need to be taken into account - and in my experience (from mapping in more temperate areas) surface more detailed than paved/unpaved rarely can be added from aerial. Unless it is extremely good with resolution measured in centimeters.
False and meaningless comparison: No one is arguing for =paved + material=asphalt , or =unpaved + material==dirt / =laterite
“Wikipedia’s dirt-road article recognises laterite/murram as a distinct subcategory, not equivalent to generic dirt” doesn’t explain the data structure. Can be read as favoring =dirt + *material=laterite
Practically speaking, it all boils down to how you want routers which don’t implement explicit support for laterite to treat paths like that:
a) fallback to the more generic (dirt/clay) surface without considering the subtag
or
b) treat it as an unknown surface
both variants can have pros and cons.
Historical momentum. Maybe it should have been surface=paved + material=asphalt/.../..., there are certainly good arguments for a hierarchical scheme like that with gradual levels of detail.
(This is where the OSM tagging model really shows its shortcomings.
In e.g. wikidata this wouldn’t be an issue. Data consumers would still be able to see the ‘laterite is a subclass of dirt’ → ‘dirt is a subclass of unpaved’ → etc… property relations and simply walk up the chain and handle the most specific item they know about. There’s no way to encode this property relation with OSM tags.)
I would rather focus on whether typical mappers would consider laterite a subtype of dirt/soil or whether it would be considered as distinct surface type.
Better framing, adopted. The proposal now leads with permanence: “surface=mud: for places that are wet nearly all year and never or rarely dry” and adds that surface=mud applies only in the rare case where a laterite surface stays muddy year-round and never dries.
Agreed. The aerial caveat has been strengthened: colour alone is insufficient even in tropical areas; ground-level photography or confirmed local knowledge is required to confirm the material.
The field survey is clear: 37 of 39 confirming respondents used distinct category names, murram, tany mena, din daeng, red road. Two used “dirt road” as a colloquial local label (South Africa, Madagascar) but both confirmed the same distinctive wet-season behaviour and agreed the tag is appropriate. The behaviour is what makes it a distinct surface type, and the data is overwhelmingly in one direction.
One additional response received during RfC (Colombia, second independent confirmation of beige/orange roads with wet-season plasticity). Respondent counts updated to 42 responded, 39 confirmed.
@julcnx : kudos on this well-thought-out proposal , and especially on the way -if I understand correctly- you solicited feedback from around the world, including from 100 local mappers in communities that may not be as active in the global discussion threads.
Out of curiosity—and also because the only specific subtropical road surface I’ve personally encountered is sand in the Sahara dunes:
is laterite also what western travellers when being abroad sometimes refer to as “peanut butter mud” ?
and would an example as below qualify as surface=laterite based on the criteria of colour and sticky muddiness, or do you need to observe the cracked dry state as well ?
I know I’m jumping the process here, but I really want to add language along the lines of:
Routers are encouraged to query historical weather data for the area and consider surface=laterite impassable by wheeled vehicles if the way area has rained in the past [X] days.
As I understand it, one of the defining characteristics of laterite is a two-stage response to water: initially, it remains hard but becomes incredibly slippery, and after extended exposure, it becomes soft, sticky, and nearly impassible.
I just hope this proposal will not fail due to too many people trying to enforce something truly untypical for osm tagging, that is to try and mimic e.g. HS-codes and their deep (contrived) hierarchies. That certainly would be the wrong reason. The only valid reason in my opinion would be: This will forever be niche. And that is ruled out, by what I read in the proposal.
In osm this would be quite easy: surface=dirt; dirt=laterite, along the lines of highway=service; service=driveway.
OTOH, to know, that when encountering a highway with surface=dirt it would be great to check for material=*, a consumer would have to hold a table of what other attributes may contain further details.
BTW: Has somebody here ever heard or read of a highway made out of some material? I am not a native English speaker, but I do not think I ever heard or read that. I am familiar with “paved with asphalt” or “surfaced with gravel”. But made out of?
A boardwalk or duckboards could be made of boards/timber/planks/etc. A path might be made with pebblestones. A grating might be made from steel bars. A bridge would perhaps more typically be built out of something rather than made, and here the named material might not be the one used for the surface.
Let us focus on the proposal: One more valid (in my opinion) opposing voice might come from the this is not something a mapper can distinguish on the ground camp.
PS: You are the one to post the Finnish tracks guide? We also have lots of moraine, it has been compacted by 900m of gravel respectively 2500m of ice during the last 100k years, and the composition of the material lends itself very favourably to create a surface=compacted, after sieving out the cobbles or boulders, that is. All the while we do not map surface=moraine, nor do we map surface=ground; material=moraine which would be the ultimate in correctness. Still I do consider the laterite problem not the same.
Yeah except that was about public roads, not tracks.
I have not been involved with the local community for very long, but my impression has been that most people are more comfortable doing their best to use pre-defined tags and values, while this level of detail hasn’t been of interest to many. Finland still has massive areas that lack landuse, and public roads that don’t have even paved/unpaved distinction tagged so my personal opinion tends towards mapping things that would be directly useful for general routing and map rendering. Who cares about moraine, basically.
I don’t have personal experience with laterite, but from the description it seems to more than meet my criteria for usefulness. Even considering everyone is free to map what they like, it’s pretty wild that we don’t even have an established value for something like this, while elsewhere people are tagging the species of individual trees.
