new armchair mapping in Thailand

Hello. I would like to do some armchair mapping in Thailand, but of course I would like to avoid any problems like mentioned here. Is there aerial imagery available from which I can map simple features like roads, landuses etc? Also, what are common mistakes that I should avoid?

Roads were traced by Facebook and others (often low quality) - there are not many left.
Landuses: most are still missing. Differentiating between (natural) forests and plantations (oil palm, rubber trees) can be complicated.
Buildings were traced by some beginners in HOT tasks (see above).
Waterways (river, riverbank, stream) are also missing; take care of severely reduced waterflow during dry season; also the course does change quite often with not so wide rivers…
Take care for alignment of the imagery - and unfortunately, there are not so many GPS traces in the province.

Thank you for the advice. I’ll take that into account when I map.

Hello Friendly_Ghost,

welcome to Thailand. I have split off your posting in a new topic, as it is not related to the previous one.

As you already have a few hundred edits, I suppose you are already familiar with the editors. So especially when creating building outlines make sure you get the share right. In your JOSM it is the “Q” key to square buildings. Also when adding, double clicking “a” enables the angle mode which allows to easily draw at fix angles.

You will find plenty of areas where you can still add new features from aerial imagery. As others pointed out, keep an eye on alignment of the imagery and try to double-check. If the alignment between various sources changes, be extra careful.

Also try to avoid guessing too much. The more guessing what the feature on the ground could be, the more like it is to make mistakes.

Rivers are very under-mapped. As most bridges have a sign telling the name of the river, I usually create a small stub under the bridge with the name. Others do similar. You could extend these rivers in both directions. On forks continue without naming the river, as you can’t tell which on carries on the name.

You could also help in reviewing existing edits and fix the worst mistakes.
I run the osmose backend for Asia, including Thailand.

You can enter using Select the problem category you think you could handle from remote. Unconnected ways can be tricky, as housing estates could have private roads ending at a wall not visible on aerial.

Some parts of Thailand have street-level imagery in mapillary. You can use this to confirm. Otherwise leave the ambiguous cases unresolved.

Always feel free to ask here in the Forum for advise if something is not clear.


I’m sure plenty of general advice for newcomers has been written, some of which you may have already seen. (You might want to check out this thread from the help site, though it’s quite old and the editor landscape has since changed.)

Off the top of my head, some simple-yet-common mistakes (my experience is mostly in urban areas) include mistaking straight lines (e.g. canals, walls, simple gaps between buildings) for roads, connecting a street to a road when it’s actually a dead end (with a separating barrier), and tagging private driveways as public roads. Getting bridges and layers right can also be quite tricky for newcomers. One of the more frustrating things to see is buildings traced with a lack of common sense, resulting in horrible parallax errors when the imagery is at an angle. In the below picture, you’d want to trace the building outline at its base, as in the green square. You wouldn’t believe how often people trace the red line instead (or, less badly though still a problem, trace the roof outline instead of the base).

I’ll try to think of some Thailand-specific tips—there’s a lot of stuff that’s obvious to locals (like telling the brand of a fuel station by the shape of its roof) that others would have a hard time recognising. Problem is, this sort of background knowledge is hard to identify and write down without some sort of triggering question.

Hello again. My apologies for replying a bit late. I have made a couple of changesets now. I think I’m getting the hang of this. Thanks a lot for all the good advice.

Concerning rivers, I discovered that someone merged an administrative boundary and a river in this area:

Yes, on occasion creating a relation and sharing the way is a faster alternative than retracing an entire river. I’ve done it a few times myself. I do much of my mapping in rural Alaska where boundaries often follow the path of a river or stream. It’s not illegal but does add extra complexity to both objects, especially for mappers who aren’t familiar with relations.

In the case you cited, the way is shared by the river and four different boundaries making it a tricky beast to edit. So if any nodes are moved, for example, to better follow the river’s actual path, all the boundaries are affected similarly.

Best not to fool around with something like that until you have more experience.



As Dave said, it can have advantages. Often boundaries are defined as following the middle line of a river. Having the boundary attached to the river ensures that it always follows the river in case the geometry is improved.

In a lot of other cases I would advise against it. It adds extra complexity. So to make life easy for all, do not glue landuse or building polygons to other features.

They also belong to slightly different levels of abstraction in our data model. Landuse polygons abstract an area, while linestrings only get their width by explicit tags or by implicit assumptions of data consumers.
The linestring geometry like roads is also used to for routing graphs.

Do not worry about “unmapped” space too much. It is very easy to end up mapping for the renderer, to “draw” a nice map in the carto rendering for zoom 19+.

I’m still a bit skeptical about the river/boundary thing because of the added complexity, but I’m not about to touch administrative boundaries, so I guess it’s fine for now.

since I’m mapping remotely, my efforts are limited to landuses, waterways, roads and buildings, so that’s what I’ve done so far. I 'm wondering how valuable this data is.