How we use highway=track in Thailand

How do we intend to use highway=track in Thailand?

My experience is mostly based on the usage in Europe/Germany.
I am used to have this tag for roads that are primary used for agricultural/forest use. So the road the farmer takes with his heavy machinery to reach the fields is a track. Same the roads in a forest used to get wood out.

They are often unpaved like this:

in case the way is used more frequently or with heavy machines they can be paved as well:

The important part is their usage/function, not the surface.

A track is not used for roads that provide access to villages. These roads that connect villages to other villages or to the major road network are tagged as highway=unclassified.

This tagging based on usage/function is also important as routing engines can base their routing decision on it. They often do not consider tracks when calculating a route for cars.

I suggest we follow this practice also for the tagging in Thailand.

Roads connecting villages/hamlets will be at least a highway=unclassified, never a track. The road villagers use to get to the next major highway will be at least a highway=unclassified, never a track.

Track will be used for roads that serve agricultural purposes. So highways next to the fields could be this. A road next to an irrigation canal that does not serve to connect villages would be a highway=track.

Shutterstock example for a highway=track

Shutterstock example for a highway=track

By specifying the surface (using surface=* and tracktype=) and width (using lanes= and width=*) of the track we can give more information whether this road might be useful for other routing. Think of a routing engine for bicycle or motorbikes.

When tracing roads from aerial images it would be important to check whether it connects houses to the road network hinting for a unclassified or whether there are no buildings visible indicating a track.

Thanks for the post, Stephan.

I’ve wondered about this for some time, and tend to tag a track as being a road most people would not want to drive on. That should satisfy the routing issue. Maybe the locals do, in good weather. Very subjective I know, but I think of map users who may resent being sent down a narrow muddy lane.

I agree that if a road is unpaved but good enough for most weather and actually goes somewhere, it should not be a track.

BTW, a lot of the ‘tracks’ you show in the 3rd and 4th photos are being paved around here and often connect villages. I tend to call them ‘residential’ if I can find a few houses along the way.

The “track” tag is not as clear in south east asia as it is in Europe: my impression is that motor vehicles may be used there by anyone, not simply for agriculture/forestry. So I tend to emphasize the purpose (usage/function…) of the way. Unpaved “public” roads are not rare, at least on the level of residential/service/unclassified; I’d suggest to add a tracktype tag for such roads, too.

When working on aerial imagery from suburban areas, I quite often come across what are physically dirt tracks leading through patches of grass or former paddy fields, but which now link to some houses. I tag these as highway=residential surface=unpaved, but it’s quite clear they previously served as agricultural tracks. I get the logic of distinguishing by function, but I can imagine cases where we’d end up trying to determine whether the building at the end of the path is a residential home or a farmhouse.

So we are all coming to the conclusion, that roads connecting villages are never “highway=track”, right?

highways with the clear function to interconnect are at least unclassified, roads to service houses but not used for interconnect are residentials?

So maybe easy things first.

This is an example for at least a highway=unclassified. Could also be a tertiary road but can’t tell from an aerial. Load in your editor and check the aerials:

This area is more complicated. Are these residentials? Or unclassifieds? I would favor the later. A comparable setup is already properly tagged and consists of tertiary roads with proper refs (

What about the (as yet unmapped) paths leading into each of the orchards? Should they be highway=track, since orchards are agricultural, or highway=residential, as they are used by the people who live there?

Reading the above, I dont think we are all in perfect synch with each other…
In principle, I agree Stephans broad view that unsurfaced ways connecting villages, are highway=unclassifed, surface=unpaved/dirt, (in short, a UPR) and for armchair mapping, I go by this.

However, myself and Chris P. probably add more GPS surveyed (offroad) ways than most, and while we do look at the places they connect, we also rely heavily on suitability as a guide. The definition needs enhancing but briefly :

Unpaved Road - You could drive a normal 2 wd car down, and is of sufficient width to allow it. Smooth gravel, and packed dirt would be typical surfaces. If the odd rut makes passage slow, but still attainable, it stays as UPR. If the ruts and washouts are too excessive to allow unhindered travel, then it would likely be tagged as a track instead.

Track - This is where the way is only passable by a 4x4 type vehicles, offroad-motorcycles, tractors etc. Some of the mountain hamlets and settlements are connected by these ways, but in this case we dont tag as an UPR just because it connects villages.

Path - This is a narrow width trail also called “Single Track”, where by inference, it contains sections that are too narrow for a 4x4, or just too steep/rough to drive a vehicle along. Off-road motorcycles can be used without difficulty, and while it may not be impossible to get a Off-road buggy, or winch equipped 4x4 through, the actual usage would indicate this is not happening in reality.

Footpath - This is reserved for places where vehicles are not allowed (not many in Thai) or more importantly, where the way is too steep/narrow/rough to allow passage on a motorbike without danger or physical manhandling of the machine. We also use this where we forge paths through jungle and the way is not defined well on the ground.

This may not be perfect and open to argument, but its how we tag when we ride these ways… as I have seen somewhere on OSM before … if it looks like a duck, quacks like a duck, then tag it as a duck ! Tagging what you see is far easier that trying to tag by guessing a use… Paul’s example of an orchard track is very apt.

I am aware of tracktype, but frankly it takes enough time to remember the basics without the refinements, given that the conditions change dramatically within just one few km length of track. We now do try to document the first change from paved to unpaved (and vice versa), but dont attempt when there are frequent changes, such as in the mountains where the steep bits gets some concrete, and the level bits, not.

Hope my pennys-worth helps, Russ.

Yep, how Russ describes is basically how I do it.

‘The important part is their usage/function, not the surface.’

Not for those actually looking at an OSM map (or using one for app routing) trying to navigate terrain.

‘By specifying the surface (using surface=* and tracktype=) and width (using lanes= and width=*) of the track we can give more information whether this road might be useful for other routing. Think of a routing engine for bicycle or motorbikes.’

These don’t change the visual display in any useful or usable way. (At least, not in any OSM data displaying maps I’ve seen.) Yes, it might aid routing apps but they are all but useless (possibly troublesomely/dangerously so) when travelling away from major roads and highways. For that reason, I don’t even think to complete these tags for paths, tracks, minor roads… (Russ is far better with adding such details.)

I typically use ‘Minor Road’ for roads whether paved or unpaved - as long as in relatively fair condition - that connect from a major road to a village. Not all - if it’s rough going then I’ll label as ‘Track’ - similar to Russ. I use ‘Residential Road’ for fair condition roads within close proximity to a village (residential roads being in residential areas). By ‘relatively’ I refer to the local environment, terrain and general condition of ways. This is something that depend on what area you’re in… Eg, close to a city you might expect roads in a certain condition, far out in the country you might expect quite different.

I must say, not at all to knock the effort you guys put into mapping, but when you trace from sat’ photos, you really are playing a guessing game on what condition the paths, tracks, minor roads are like in the country - or if they are even there anymore. There’s not just the matter of clearly seeing sufficient detail to classify accurately, there’s also the fact that, in the country, routes and ways can change often over years (or even a year) - in use, direction and standard - and so the age of the sat’ pics you’re using can be an issue too. (Eg. rainy season makes a big impact every year.)

As a keen motorbike trail rider and sometimes mountain biker I have spent quite a lot of time trying to spot new routes from sat’ pics (paths and tracks mostly) and then gone out to check them. So, I consider I do have a fair understanding on this issue. I don’t map anything onto OSM until I’ve been there for this reason. Obviously this isn’t a method everyone else mapping to OSM can (or should) practically follow, but I’d just advise caution in trying to classify a way in the country in too much detail just from looking at sat’ pics - as in my experience, that can’t be done with accuracy.

There can be a difference in opinion on what the key use of OSM data is. For me it’s navigation. Classing any other feature as being more important then surface condition, for the very means by which navigation is permitted, is a mistake, surely - as it’s so critical. To be blunt, if ways are classified with a priority to function over surface, OSM would be a poor (read often troublesome and sometimes dangerous) map for navigating country routes. I can understand wanting to organise way type according to function - and it is important and useful - but definitely not more so then surface, in my opinion.

‘Roads connecting villages/hamlets will be at least a highway=unclassified, never a track. The road villagers use to get to the next major highway will be at least a highway=unclassified, never a track.’

But what if there aren’t any Highways connecting - they are tracks or even just paths! (I know plenty. Sorry but some parts of Thailand’s infrastructure might still not be as developed as you would like to think.) Obvious definitions for way types easily judged by visual appearance are being overlooked - for the sake of fitting in with some derived ‘organise by function’ schema… It seems this might be due to a wish to cater for routing apps - nice theory but in practise, in the countryside of North Thailand anyway, little chance of getting safe results. (Unless everyone becomes a good dirt bike rider!)

When Russ is up CR in a few days we’ll take some pics of ways leading to villages and hamlets - and between them - post them here and see if people think they should be called Highways or Tracks…

Note, what’s crucial here is the way OSM follows typical mapping protocol by using thin and dashed lines, etc. to represent smaller, poorer quality surfaced ways… Using Highway tags (for what people would typically look at and call a track or path) that are represented by thicker solid lines on all the OSM derived maps I’ve seen - due to a ‘function more important than surface’ opinion goes totally against that established and recognised method. (Check the path and road type graphic icons used on Potlatch, for another example.)

Some of you guys should get out in the country and try using a map to navigate. See what land and way conditions are like here (it’s not Germany). I see too much of a blinkered approach focusing data integrity and structuring concerns around unrealistic objectives - such as app routing in undeveloped country areas away from major infrastructure (roads). A maps ability to help a user (navigator) visually distinguish a ways (surface) condition is key when attempting to travel over undeveloped country terrain…

Here’s pics of the main 2 tracks going into and through the village Ban Ra Bue, Chiang Rai… (I’d class this as a village and not a hamlet - it is of a fair size and has a 1 or 2 shops, etc.)

Note, these pics were taken today and show the tracks after rainy season repairs. During rainy season, these tracks are often impassable by a pickup due to deep ruts and washouts. I know various other villages and hamlets around with access only by tracks like this - some even narrower and in worse condition. If you want more pictures, let me know.

And note, from the Wiki for minor road / highway=unclassified:

‘Unclassified roads are considered usable by motor cars.’

‘…should be tagged as highway=track if they are unfit for standard motorcar usage but available to 4x4, bikes and foot usage.’

So, only if the way could be transported by a standard motorcar should it be tagged as a minor road. If a pickup, offroad 4WD or similar is required then it should be tagged as a Track (if wide enough - if not then use Path). TBH, a few of what I classify as minor roads really shouldn’t be - but I am considering the typical way condition in Thailand’s countryside and the fact that many vehicles used in the countryside are pickups.

I don’t think either Track or Path should be routable for motor vehicles - unless ‘offroad 4WD’ or ‘dirt bike’ options are available. (I don’t know routing apps/scripts - never used.)

Hello Chris,

in your example the road is quite bad, but still it is the connection to the outside world to probably hundreds of people in that village. Perfect to be qualified as an highway=unclassified.
I understand it might trick someone just looking at the mapnik style of OSM to think they could use their city car to drive up there. But this is an issue with the renderer not using attributes available in OSM.
The 4044 which the current track connects to is also currently tagged as unclassified. Maybe worth upgrading to a tertiary given its role to collect traffic from a bunch of villages and leading them to higher category roads.

Leaving away the tagging of the surface tag just “because no one evaluates it” is in my opinion the wrong way. You make it even worse this way. Once the main renderers (we’re talking about them. I could create you a map style in 5 minutes taking into account the surface tag) start evaluating it they rely on it being there.

Your quote of the wiki page is misleading. You forgot to quote the beginning of the sentence where it limits it’s definition to the UK: “UK: unsurfaced country roads (“green lanes”) should be tagged as highway=track…”. Given the typical road quality standard in the uk this makes perfect sense.
The wikipedia definition of a green lane also suggests that it is something different than the main road leading to a village: “In particular, a green lane is unsurfaced, and may be so infrequently used that there is no wearing of the surface, allowing vegetation to colonise freely, hence “green”.”.

Even more, refer to the definition of track where they also link to an issue with the map style explaining why tagging them as a track is actually tagging for the renderer and support for evaluation of the surface tag by the renderer is requested:
"Do not use tracks to represent public unpaved roads in built-up areas [1], that would be considered tagging for the renderer. In this situation, classify them as usual according to the conventions in your country, and also provide a surface= tag."*

As you also noted in your last post a highway=track should not be routable for cars. This is exactly the problem. This leaves all those villages unreachable for routing.

‘Green Lanes’ is one term used to describe (often generically) ‘UK: unsurfaced country roads’ - it doesn’t mean they’re all actually the Wikipedia definition of a ‘green lane’ - like, in the UK, ‘let’s go green-laning’ means lets go ride dirt bikes on dirt tracks… It’s difficult to stop things turning green in the UK - not quite the same as Thailand… A lot of ‘green lanes’ can be very brown and muddy - I know as I grew up surrounded by them. But, yes, I missed out an important part there - ‘UK: unsurfaced country roads’. Like all of them - which is more defining - it doesn’t state only specifically actual ‘green lanes’.

In the UK I doubt if you could now find a village/hamlet only accessible by what I would call a track… A few decades ago, you probably could. But this is Thailand and the countryside doesn’t exactly compare to development in the UK’s countryside.

The Wiki infers to me that Unclassified roads are considered usable by motor cars - standard motor cars. If not, and wide enough then it’s a Track. Other posters on this thread have said similar. Looking at the ways in the pics - I’d call them tracks and not a road…

Yes, you’re right about tagging the surface - and I will go back and retag all the trails I’ve added at some point… Too busy out actually mapping them as that’s far more fun - and I really wasn’t thinking much about renderers then - just what Track and Path and Road mean by definition… Though, Track and Path could just be defaulted to unsurfaced when running mkgmap (or by routing programs) - and those are the ways I map most. (Not that all routing apps and devices pay attention to such tags anyway.)

I’m not sure of the relevance of quoting on tagging in built-up areas as a way of tagging for the renderer… I’m not mapping built up areas - if I was then you probably wouldn’t find tracks, though. Like I and others have said, an unsurfaced road isn’t necessarily a track - only if its condition is too poor for a standard motorcar to use.

Re. your last comment - yes, exactly… These villages and hamlets with poor track access only should not be available for routing to standard users (in standard cars) - or problems are bound to happen. You can not insist that used renderers show or routing scripts detect the difference between way and surface types on any map app / device. Some do, some don’t. Great if they do and there’s a clear option to turn on (default off) routing for offroad bikes and vehicles but otherwise…

Do note, for any village or hamlet where access is by an unsurfaced way I see as OK (sometimes vaguely) for a standard motorcar - I always tag that way as Unclassified. I agree that defining function is important… Up to a point - being when that can make navigation misleading/difficult/hazardous. It’s not like users won’t be able to see these few (most are accessible by Unclassified) hamlets/villages - just they’ll only be able to route to the nearest point on a road of fair condition for a standard car (and driver - which makes lots of difference - even if they have the right kind of vehicle.) They can still look at the map and navigate - hopefully with more caution and not just blindly following app routing.

Also, it may not be the case that people would just be completing a short, worse, section at the end of their route to arrive at their destination. They could be routed through long sections of track as a shortcut between more major roads - as these villages are often situated. (I think most people are aware that GPS device apps are notorious for this - even in developed, well mapped countries.) Eg. this track - which I’d guess you might like to label as Unclassified. You really don’t want to try this track in a standard motorcar, for sure:

And I refer back to this quote from the Unclassified Wiki page (which surely no one can see as misleading):

‘Unclassified roads are considered usable by motor cars.’

It doesn’t specify what kind of motorcar so that defines any motorcar, i.e… including standard motorcars. (e.g. 2wd without high clearance and with road tyres.) No way you can drive what I label as Track in a standard motorcar without problems - even after rainy season repairs.

JOSMd surface and vehicle tags for my ways… Will check other editors tagging as and when I come across it.


Any Path is now:
Surface=unpaved (unless already set to something different)

Any Track is now:
Surface=unpaved (unless already set to something different)

A Track might still be passable by a standard motorcar. (Unclassified HAS to be passable by a standard motorcar.) So, can’t have Motorcar=no for all. Tag types that can be used to show more detail on condition are Tracktype, Surface, Smoothness and 4wd_only. These’s some overlapping here but I see Smoothness and 4d_only as being most practical - though 4wd_only should be renamed as ‘offroad vehicle only’. As I have no idea how routing scripts would get on best with all of these tags I hope the Surface tag will take care of things well enough and that Path and Track are excluded by scripts for standard user routing purposes - which will all depend on routing app designers as they’ll be able to include or exclude ways and tags as they like, I guess. At least it will be possible for users to see visual differences between what is actually a path, track and road - as long as data is rendered with a difference - in a style fitting with how different map ways are typically drawn.

Footways and Cycleways are now all Motorvehicle=no

For Unclassified, I checked through all associated with my user ID and have selected either unpaved or paved as I recall. JOSM tells me the Motorvehicle tag is not required for Unclassified.

Other tags I’ve left as is…

I covered the area roughly bordered by the 107, 1, 118, 1150 as that’s where I map - and I have a knowledge of the majority of country ways in that area.

Let’s look at another non-european place, India. A few years ago I travelled to Nelliampatthy mountains. On my bicycle, I could go on roads of different qualities. Sometimes I happened to reach guest houses at the end of terrible unpaved ways (worse than the roads on Chris’ pics). And some worse than

(Bigger version:
But the owners told me that people do get their by their normal cars (not 4wd). I think I could do so with my car also, bit I wouldn’t want to…
I understand the issue with the maps. Though it is not so complicated to render such roads different from “normal” roads, most renderers don’t. I often requested to differentiate between paved and unpaved roads, but the map makers live in modern parts of Europe and don’t bother about such strange exceptions (though there are unpaved roads in Germany).
If we do not get things usable for asian countries, why should peolpe there use OSM maps, and eventually contribute to OSM?

Look - it’s there on the Wiki…

‘Unclassified roads are considered usable by motor cars.’ (meaning standard cars)

I am not an arm chair mapper. (Not that that’s a bad thing.) I am out in isolated countryside once or more a week. (Unless it’s too bad from rainy season - I’m not that mad.) I ride dirt bikes and have driven 4WD pickups over this terrain. If I see a way as passable by a standard motorcar, it’s wide-ish and it’s function suits, I label it as Unclassified. If you need an offroad (meaning not on a ‘road’) vehicle and/or narrow then it’s a track. That fits with the OSM Wiki way definitions. My primary care isn’t how stuff renders - it’s about the definition (a Path is a path, a Track is track, a Road is a road). The reason I care more for the definition is I care about map user (navigator) safety and them avoiding problems, hazards and dangers.

This is a classic surveyors vs IS office argument! (I have worked both sides on GIS services projects.) If you want to have a meeting and try drive a standard car on any of the ways I label as track then come to CR, rent one in your name and I’ll pay… You can cover for the damages. (Or bring your own if you’re feeling risqué.) It’s likely some will find the one’s I label as Unclassified difficult even in a 4WD pickup - unless they’re experienced rough terrain drivers. If you think I’m underestimating what a standard car can travel over - LOL. I have never seen the locals on these ways in standard cars - only pickups - but you can ask them if you like.

What do you mean about getting things usable? You think a map that is misleading is usable? Or an app/device that routes you on unsuitable (possibly dangerous) ways is usable? One that you can visually be informed by accurately of way surface condition is usable. One that routes safety is usable…

Seems like there’s a few with a strange desire to have anything called hamlet and village routable by standard car - regardless of bad way condition… Do you all work for a motor recovery and repair service?

EDIT: Seriously, if any of you with doubts/desires want to come up CR, I will take you to these ways and you can see for yourselves. This time of year you’ll be (relatively) OK on a moped, if you’re very fit, MTB (great trails!) - or I can strap you down in the back of the pickup. Alternatively, you can get locations and nav there by OSM yourselves now most ways are mapped around here. (But I’d advise only doing this if properly prepared.)

I fully understand that there is a difference in the road quality observed in Europe and in some undeveloped areas of Thailand.
The problem is that OSM tries to establish a basic set of tags that can be understood worldwide.

There had been similar issues with tagging of dirt roads as tracks in Africa where these had been actually major highways.
This is why I emphasize so much on the function of a road.

If a road serves as the main road into a village, it is a main road. So it deserves a higher category than a track or a path.

The definition of a standard car is depending on where you live. A standard car in Germany is something different than a standard car in Africa and something else in the rural areas of Thailand.
No one in that village you mentioned will drive a Toyota Vios. It is not the standard there. People might drive a Hilux.

With such cars the road is much more driveable. Maps are used in a context. So people will unlikely try to use a city car to drive up to such remote places.

More than that the missing evaluation of the surface tag is still a problem with the renderer. Please do not tag in a specific way just because the render result is not the way you want to have it.
Comment on the following issue to move it forward…

Yes, there’s major highways in Africa. Australia, wherever that are dirt. These are roads as they are typically graded dirt and wide - not narrow and heavily rutted.

‘The definition of a standard car is depending on where you live. A standard car in Germany is something different than a standard car in Africa and something else in the rural areas of Thailand.
No one in that village you mentioned will drive a Toyota Vios. It is not the standard there. People might drive a Hilux.’

Now you are really trying to stretch things… So you are wanting to tailor routing and OSM just for the locals? They’re going to want to use routing around where they live? So in Muang Chiang Rai the typical car is a pickup? Does the 10-20k to these ways mean that all the Vios and similar you see in town are no longer standard to the area? Who are you trying to cater OSM for? For a local to these areas, yes the standard car is likely an offroad pickup - but for the users of Thai OSM, a standard car is going to be a 2wd with road tyres and average clearance.

‘More than that the missing evaluation of the surface tag is still a problem with the renderer. Please do not tag in a specific way just because the render result is not the way you want to have it.
Comment on the following issue to move it forward… … issues/110’

Don’t know what you mean here… I’m not tagging anything so it renders how I want. What do you mean? I just went and added surface tags to all the Paths, Tracks and Unclassified I’ve edited - accurately - to give more useful detail. You requested that. Are you saying I’ve labelled them unpaved when they are paved??? All the Path and Track I’ve added are unpaved - with a smoothness of at least ‘very bad’ - if I was going to add that tag. I spent a long time going through all Unclassified I edited or added and labelling paved or unpaved as they are… Do you think I would have had much fun riding my dirt bike around mapping paved ways? Have you ever been out in isolated country (ever?) and found paved paths and tracks?

And what relevance has that issue you linked to got to this? Look at the street layout in the 1st pic - that is clearly a developed area. No, they shouldn’t be classed as tracks. I’m not talking about ways in developed areas!

You don’t want to listen to opinions of the people who actually spend lots of time on the ways you are discussing. Others have said the same thing - if they wouldn’t take a normal road car on it then they label it as Track. The Wiki says the same thing. Is this an arm-chair mappers oligarchy or something? I really wonder what your aim is here…

If I get time tomorrow I will go to a hamlet and take pics of access to it. If you can call that a road and not a track then your reasons for defining as such must be driven by a peculiar motive. You are unwilling to consider that, for the (relatively) few villages and hamlets in isolated country that have very poor access, the only access is by what should be defined as a track not a road - simply because that doesn’t fit with some rule you want to apply.

Chris. if even local people don’t use those roads with normal cars, then the situation is really “special”. I haven’t been to Chiang Rai region yet, in the far South of Thailand or mainland Malaysia I did not find such roads, so I looked at my experience from India. Then I must conclude that you are talking about even worse roads. In Indonesia, there are villages which cannot be accessed by car (also not by 4wd) - only by motorcycle on a “road as narrow as a hand” (“jalan setapak”; i.e. highway=path).

The tag 4wd_only=yes (see ) could be useful here. But on that wiki page, they suggest to combine it with common highway tags (like unclassified, tertiary etc).
Could such tag be useful in Chiang Rai’s countryside, perhaps with some additional tags?

I haven’t come across any settlements that are only accessible by path - plenty of houses though. You might still find them in even more isolated areas such as Nan, Mae Hong Son or Omkoi maybe - and I know not far over the border into Burma, you still find them. These villages and hamlets are inveriably ‘hill tribe’ - eg. Karen, Akha, Lahu, Lisu. (I wonder, if I did find a hamlet or village only accessible by path, if certain people would want to tag that path as a road too - just to fit with some ‘function over surface’ rule?)

I would say the Smoothness tag is the most comprehensive tag to describe way condition in terms of navigability - I don’t see the Tracktype tag being as much help here in Thailand. If the 4wd_only tag was renamed as ‘offroad vehicle only’ then that would make it better, however, I doubt whether routing scripts are going to care much about either tag. Surface=paved/unpaved is likely about it, if you’re lucky. (Eg, just tried - a popular GPS and nav app - and it doesn’t care about way type or relevant tags it seems.)

I think the primary importance when discussing this issue is that a path is tagged as a Path, a track is tagged as a Track, and a road is tagged as a Road - and these definitions are governed, as is established and historical, by appearance and condition - not an invented redefinition due to function…