Thailand abused as a training ground of OSM training

Did some googling, and Alan Ziegler appears to be this professor at the National University of Singapore. They did a field studies class in Khon Kaen last June/July. His contact info is here.

Just for info - the DWG have just had a message from someone involved in the work here. We’ll be in contact with them shortly…

Hi all,

We (the DWG) have been in contact with the person who organised this field trip (it was actually a remote mapping day followed by a trip to the area) and they’ve said that if they do this sort of activity again the’ll definitely add in a validation step. I explained a bit about how “validation” in OSM currently works, both on the local level with individual mappers talking to each other and with more organised remote mapping (e.g. HOT tasks).

That still leaves the question of the data that has been entered here - while some of it (e.g. the volcano) is presumably rubbish some (the field areas presumably aren’t. The buildings named “unknown” obviously aren’t called “unknown” and may not be the correct shape but do seem to exist.

Best Regards,

Andy

Found that this one is still unresolved.

The points from Andy that buildings actually exists are valid. How much do we value the existence of buildings in a vuilage above the crappy geometry of this mapping?

Is it really considered worth to keep these shapes? Otherwise generously simplifying the fields geometry and deleting the buildings manually keeping the stuff worth sounds reasonable. No one from that university ever came back to fix the mess.

Please comment either for keeping the details or removing.
With time passed now the removal should be a manual process to only remove the bad things.

A few months ago, I did sort out a few villages as I stumbled across them - the majority involved poorly mapped buildings, with just an area=yes tag. Would take about 10 mins per village to highlight the lot, then change the tag to building=yes.
Then there were a few where they had tried to draw the buildings without aligning the map, resulting in either buildings over roads, or jagged roads where they moved things around, unsuccessfully. Other errors include cut & paste mistakes where buildings have been redrawn twice.

My suggestion is if you can run a script to delete all objects in Thailand that just have the area=yes tag, then we can see whats left. I cant think of anything that would legitimately carry only that tag with nothing else.

To do manually could take a lot of time, unless you want to maybe delete on a user by user basis. I seem to recall the bad plotting was attributable to work done at one Issan Universities, and despite attempts to contact them, they never responded. (See the other post at https://forum.openstreetmap.org/viewtopic.php?id=65790on this matter).

Russ.

OMG. There is much more. I couldn’t run the full query here, but it seems to be in the ballpark of 1 million ways having only area=yes.
Looks like the biggest amount of them is in an failed attempt to map buildings in December 2018.

https://overpass-turbo.eu/s/RsG

The mass-retagging of everything as a building mentioned before might have been a failed attempt to fix this.

I added a comment here, but have not much hope of getting a reply.

https://www.openstreetmap.org/changeset/65730274#map=16/16.2074/103.1351&layers=D

The geometry only remotely correlates with the shape of the buildings.
As it has all such low quality. I would agree with Russ to just remove it. The effort to fix it is the same or higher than simply mapping it again properly.

Did anybody else receive feedback regarding this mapping?

I could run a few more queries to exclude recent mapping activities, but at least the changes I have seen so far look abandoned.

Stephan

You know, we’ve always wanted some local help with the Thailand OSM effort and I’m afraid we’ve got it now. But it’s a very mixed bag. At best.

I’m seeing a lot of wilderness_huts in northern Thailand now. These simply cannot be wilderness_huts. The few I’ve been able to check are merely abandoned or unused buildings that happen to be located in a field or orchard. Here’s a typical example from user:Aruno:
name=กระท่อมในสวนลำไย (Google translation: Cottage in the Longan Garden)
name:en=abandon cottage in farm
tourism=wilderness_hut

Someone came along later (Russ?) and changed the name:en on this and many similarly tagged objects to “Field shelter” in an attempt to salvage the data but I’m not sure that’s the correct approach.

Another thing is that I keep coming across are objects that have data in the name tags that should really be descriptions. For example, I’m seeing many nodes along the 107 that are shelters serving as bus_stops, common in Thailand as we all know, and tagged with the following (but no amenity=shelter):
highway=bus_stop
name=ศาลาพักรอรถโดยสารประจำทาง (Google translate=City Hall waiting for the bus)
name:en=Cottage Bus Stop

It goes on. Many wats along the 107 have the English word “temple” appended to the name:en. Example (made up) “Wat Tham Pla temple”. The word temple is redundant and actually incorrect. In addition, the religion tag is missing on many of them.

I’ve tried to contact Aruno without success because he’s responsible for many of these objects. Besides, I’m pretty sure we wouldn’t be able to understand one another were he to answer. I’ve pretty much given up on trying to correct these sorts of errors. I’ve learned to live with them by ignoring them. After all, this is his country, not mine.

So, once again, the openness and flexibility of OSM has both a good side and a bad side. People are free to tag things as they see fit and this inevitably leads to chaos at times. I’m guilty of it, as is Russ, and probably others of our core mapping community here in Thailand. Once we start accepting the idea that OSM is available as a mapping platform to suit our personal needs, we’re in trouble

See also some other threads on “buildings”:
https://forum.openstreetmap.org/viewtopic.php?id=65790
https://forum.openstreetmap.org/viewtopic.php?id=65217
In case of doubt: let’s get rid of them!

Well I deleted a few of the worst example villages a week ago, with no comment from the mapper… Who appears to have given up anyway.

Oh, and I confess… Yup, I change the wilderness huts to field shelters as I come across them.

I don’t see any sense in waiting and would urge Stephan to delete all the area=yes ways in Thailand, without any further delay.
Its tardy and incorrect, so why should it stay there?
After that we can evaluate what’s left, and either accept if poorly drawn but correctly tagged… Or be a bit more ruthless.
Please advise when done :slight_smile:

@Stephan … you should have plenty of time on your hands now :slight_smile:
How are those area deletions coming along ? :stuck_out_tongue:
Russ.

I am continually seeing those blasted “wilderness_huts” all around Chiang Mai province. I checked the mapping of one of those responsible, user:M’Topp Ekkaraj and all of his OSM activity, 78 nodes total, were nodes tagged tourism=wilderness_hut along with names, addresses. That was the full extent of his mapping effort.

Then I did another query of all nodes tagged tourism_wilderness_hut in Thailand and came up with 566 nodes all tagged in similar fashion.

The ones I’ve checked either using satellite imagery or ground surveys are not wilderness huts. There is some potentially useful name and address data on those nodes but I’m not going to take the time to sort it out. I’ve tried contcting the responsible parties several times without success and I’ve run out of patience.

I want to delete them. Certainly those of M’Topp Ekkaraj but the whole mess of them ought to be tossed as well.

Arguments for or against?

For.

Hallo Dave,

I can understand your frustration. As the tagging is wrong, it is due to be fixed. It could be that the mapper tried to tag something else and either bad example or editor suggestions tricked them into tagging it as wilderness_huts.
You are certain it is not the typical sala at road-side, right?

So best would be to contact the individual mapper using a changeset comment. In case of no response, contact again via message. At least the mentioned user “M’Topp Ekkaraj” used maps.me. I can’t remember that I ever got a response from maps.me users.
If no response within reasonable time arrives, simply remove then the edit and leave a comment in the changeset pointing to the previous changeset discussion. As we can’t fix it, we fix the data by at least removing plainly wrong content.

To add to the previous topic:
I extracted the names and checked. Many indicate a weather shelter type of sala in the fields, like
กระท่อมกลางทุ่งนา, กระท่อมในสวนลำไย, กระท่อมในนาข้าว

others sound useless like บ้านร้าง

a tiny fraction might be intended to contain place tags, like “ม.6 บ้านดอยหล่อ ดอยหล่อ”.

I am not certain how clever it would be to simply re-tag it as place nodes without having a way to verify the data. Similar problem than the “area” tags by that malaria mapping.

I also wonder how many of these “huts” had been tagged in this context. there is a cluster in the typical HOT activity area.

See on the map: https://overpass-turbo.eu/s/ZnF

I saw another one following some roads, which could be some kind of survey. I checked the first three “huts”, but found nothing on any aerial imagery provider:
https://www.openstreetmap.org/changeset/79211165

You can find a sheet with node-ids and name tags here:
https://docs.google.com/spreadsheets/d/1wqmyiLrZFGVdjCgR5BjClmpDQ54nxvyCw5WxW4WGosk/edit?usp=sharing

@Stephan,
I’ve tried contacting user:Aruno before about his “wilderness_hut” tags but got no response. Also user:“M’Topp Ekkaraj”. I’ll create a query to pull up all creators’ names on the 566 nodes I reported on earlier and I’ll try one more time to reach them with individual private messages. I’ll report back here before proceeding to delete them. It’s more effort than I want to undertake for these worthless additions but I’ll do it.

Dave

You could focus on the Top 5 users and ignore the rest. I updated the sheet above.
As you contacted the two top users already you covered 457 out of 515 matches. The next frequent tags of wilderness_hut are “Wiwut jojo” with 16 nodes and “Russ McD” with 14 nodes. which all seem to be in version 2, so needing an additional click to check what the actual change was.

I checked some of the other users’ nodes and the situation isn’t clear cut. One of them was an actual guest_house but because the bungalows were in a grassy area, the mapper used the wilderness_hut tag incorrectly. It should have been tagged tourism=guest_house. A few were so obviously wrong I deleted them instantly.

I will concentrate on only the two aforementioned users, Aruno and M’Topp Ekkara. I will try to contact them again through the PM system and if they don’t respond, we can delete their nodes and that will clear up almost 90% of the nonsense. The ones that were edited by Russ and me were attempts to transform the original mistagged “huts” into something more useful. We can just ignore those for now.

Addition of wilderness_hut tags was a hallmark of MAPS.ME users; it seems the app misguidedly offered this as one of the basic tags. Most should probably amenity=shelter.

I cleaned one village of area=yes

My method was to use filter in JOSM

(1) ctrl+f with search “area=yes inview”
(2) press q to square all of them
(3) unselect valid geometries
(4) delete everything else
(5) ctrl+f with search “area=yes inview”
(6) replace area=yes with building=yes

I plan to continue.

Note to self: http://osmose.openstreetmap.fr/en/map/#zoom=14&lat=16.27705&lon=103.14222 is very useful

Thank you very much of helping to clean up the mess from various organized mapping activities.

A small improvement to your process could be to check that area=yes is the only tag. Query is then:

inview tags:1 area=yes

This prevents the loss of potentially useful tags. Sometimes only the name tag given in addition, but frequently it is possible to guess from the name what attribute is missing. Google translate can give hints whether something is a school or temple.

I also noticed that in Isaan we had larger imagery offsets in the past. So try to find GPS tracks nearby to align the position of the polygons.

Depending on the person adding the geometries, you might find that some are quite ok, once rectangular and offset corrected. Others had drawn around shadows or other visible imagery artifacts. Ways with 3 nodes are often badly representing buildings. Ways with 5 or more nodes tend to be some strange outline polygons.

JOSM can filter with “type:way nodes:5-” in addition to the above query for them.

A a final quality check, you can select all elements and run JOSM validator. There should be no overlapping/self-intersecting geometries.

Regarding osmose: I recently optimized the generation process. Its update process is fnished before Noon Thailand time, taking data snapshot from roughly 6 hours earlier. Updated daily.