Village huts mapped the size of football pitches

Stumbled across another area in Thailand where in the spirit of stopping Malaria, users have drawn buildings way oversize, and given them names which I suspect are more related to the Malaria project than the actual building name (H2, M16, 51, etc).
Not only that, we have a few roads and tracks sprinkled in, that don’t connect to anywhere… OK, not a crime, but when you can’t be arsed to trace the road further, cite a source, or even attempt to give it a surface tag … yup, sloppy mapping.
My understanding is that OSM is not to be used for projects that carry references to individual work ? How does the rest of the community view this?
Anyway, have left a changeset comment, and will change some buildings to areas, just to make a point - lets see if anyone shouts.
The area in Thailand is here :
No point in contacting the contributor, or asking for a block … they got bored and gave up mapping years ago, after they got a warm and rosy feeling from playing their part in saving the world :slight_smile:

Absolute garbage. Sloppy, careless and add any other negative adjective you choose.

The names on the houses are apparently the names of the owners (but who could know that for sure?) and that would be useful if it were tagged properly.


Also, did you notice the areas tagged as area=yes with a numerical description nearby. They were originally tagged as landuse=residential by the same idiot who did the houses, user:jarinya manolaeven, though they are primarily farmland or wooded areas. Then some of his buddies, ChoosakN and Elijah Filip came along and “fixed” them by removing the landuse tag and replacing it with area=yes, completely forgetting to tell us what the area actually is.

This is the future of OSM, open to any fool who can work a mouse and use a computer.

One more comment: some of those named buildings were supposedly removed or reverted by user:woodpeck_repair but they are still present. Apparently, the revert wasn’t completely successful.

I resisted the impulse to delete everything in that area so that others could see what we’re up against. The reservoir is horrible too. It might have been drawn with poorer imagery than we have now but in any case it’s a wretched job of tracing.

Yes, “นาย …” means “Mister …”. Found that in Indonesia, too: “rumah Pak Ahmad” - “House of Mister Ahmad”

I made a lengthy and somewhat heated changeset comment to one of the mappers responsible for the large areas with the numerical tags I mentioned above. User:Elijah Filip responded today:

I’m going to respond but I wanted to present this information here for advice about how to respond. If these sub-national boundaries correspond to sub-districts or some other officially defined boundary then my suggestion would be to tag them with the actual names of the areas. However, I think the best solution would be to have him simply remove them. He calls them pollution and IMO that’s what they are if left in their present state.

Opinions, suggestions?

Appears to be this changeset?

The areas are clearly muban (meaning village, not housing estate), and should probably be represented by place=hamlet. These are usually mapped as nodes, but I don’t think there’s harm in converting these already-mapped areas if there’s no redundancy. I’m guessing the boundaries were manually traced, and might not accurately reflect official lines, but then I’ve never heard of official boundaries for muban. Are they drawn at all?

The numbers in the description field appear to be geocodes similar to the system used by the Department of Provincial Administration. I’m not sure why they have ten digits, since villages (muban) are supposed to have eight, according to this blog post by Ahoerstemeier (MaewNam on OSM and here in the forums). They should be a clear fit for some designation of the ref key.


Yes, that’s the correct changeset but the tagged areas I’m talking about aren’t easily visible from it. I should have provided more info earlier.

Here are the OSM ids for three of them I located near the big reservoir. Way: ทุ่งไก้แจ้ (691844251), Way: คลองปูนริมน้ำ (691602849), and Way: ท่ีทำกินทับทิม01 (691849823). The last one is tagged landuse=residential rather than area=yes but it’s so inaccurate it’s totally worthless as drawn.

(Aside: I did some work on the reservoir outline. I haven’t added the islands yet but it looks a lot better than it did before.)

I have seen other places where arey=yes polygons are used to store internal identifiers in the OSM database. Recent example open in my editor is way 680693588.

I contacted editors multiple times, changeset comments or direct messages. I never received any reply. Not the best background experience to approach new users making the same mistakes again in these campaigns.

In case you have not yet noticed: iD stores additional details in the changeset. For example how many edits a user did so far and whether a review of the changeset was requested. Obviously in all these directed editing campaings no quality control happens, not even changesets with explicit wish for review are checked by the organizers.
It also shows how many warnings about wrong tags or overlapping ways they willingly ignored when submitting.

Given that these commercial “do good” entities like HOT try to optimize their profit it is understandable to offload these tasks to the local community and focus on more fundraising campaigns to bring in more new mappers.

To follow the process: It is absolutely crucial that you leave changeset comments in the relevant edits. This proves that we tried to resolve it. Otherwise DWG can’t help.

The vast majority of directed edits in Thailand are violating the Organised Editing Guideline and could last-resort be reverted:

A while ago I also mentioned this very prominent in the Thailand wiki page.

I would like to keep the local people involved in some of these campaigns and help them mapping properly. We need on-the-ground people.

Your message reminds me that I wanted to compile a case for DWG. I found areas where they publicly tagged malaria cases. I think it is a very sensitive personal information whether you have malaria or not. In Europe, publishing such medial detail is probably in violation of the GDPR. End even if not, I strongly consider it inappropriate and it does not belong into OSM.

As these Organised Edits clearly outnumber us and we have not the capacity, can we agree on a policy on how to deal with it?
Shall we define a process on how to approach each editor and define criteria about when to revert the changes completely?

BTW: As some of the editors have discovered the functionality to create circles, we end up with more and more silos in Thailand. Especially funny, given that these buildings are a source for created buildings in other software like the recently release MS flight simulator.

In this area you can see lots of the weather shelters on the fields are mapped as buildings. While technically correct, in the context we seen them used they clearly aim at finding inhabited places. The map is covered with tiny “buildings” only a few square meters in size in the mittle of fields. For example this one roughly 3x3 meters: way 690768512

The whole episode leaves a bad taste in my mouth. I honestly don’t know how we should deal with this sort of thing going forward.

I’m including the recent update to my changeset conversation with user:Elijah Filip here just to keep you all in that loop. Nobody commented on a solution to the areas mentioned above so I essentially told him to remove them. At least “Elija” had the courtesy to respond, unlike so many others working on similar projects.

I did suggest that they be retagged with place=hamlet though. While some of them are redundant with existing place nodes, many don’t appear to have otherwise been mapped, and it would be beneficial to retain the information in some format. I don’t see reason to believe that the village names aren’t accurate?

I’ve reached out to MaewNam for comment on the geocodes. The ISO3166-2 key is already fairly extensively used, and if these are indeed an extension of the standard they could probably by extension be tagged with a similar key.

I have no problem with waiting a bit. I have already modified some of the landuse=residential areas (unnamed ones) because they were ridiculously oversized and encompassed so much farmland and woods that I just couldn’t let them stand as is. I also “fixed” some of those huge “football pitch size” houses by adjusting them to the proper size and removing the names.

But I have plenty of other stuff to do while we’re sorting this out. Standing by…

Here is the latest changeset comment from user:Elijah Filip (emphasis added)

Thanks are due to him for responding so clearly and helpfully. Based on his response, and especially the portion I emphasized, I think the tagged areas in this changeset and others should be removed. I would like a consensus on that before requesting that he do that.

I also think he should share this thread and the changeset comments with the sponsoring organization so we can reduce theses sorts of uses of OSM going forward.

If I don’t hear any calls to keep these areas, I’ll ask him to go ahead and remove them from the OSM database.


** EDIT ** My response is below:

Hello Dave, Paul,

please follow up on this topic. Again for reference, I see that discussion also happens on this changeset:

If the names do correlate to places we could add, adding a place node at a reasonable location should preserve this detail.
I still do not fully understand where the data was coming from, whether it was an official source according to the code or hand-drawn from local offices.

It sounds that the name is actually referring to a place. So place=hamlet or place=village might match. Please check.

And as a side-note: The privacy violation I mentioned earlier is reported to DWG with reference Ticket#2020100410000051
I suggested a redaction to completely remove it, including from the full history planet dump.


Eli Filip replied to my changeset comment the other day but I forgot to add his most recent message here. (I’m traveling at the moment). Sorry for the delay.

Edit added 2020-10-13

So, what do we want to do about this? The areas I checked using the supplied Query don’t seem very useful. They don’t align well with existing OSM drawn residential areas (which are admittedly, very approximate) so my feeling is to simply remove all of them.

I edited my comment above but decided to repeat it so that people who are subscribed will get a reminder.

So, what do we want to do about this? The areas I checked using the supplied Query don’t seem very useful. They don’t align well with existing OSM drawn residential areas (which are admittedly, very approximate) so my feeling is to simply remove all of them.

Hello Dave,

I thought about retaining the information from the name tags. The area itself looks very unofficial. This is nothing we could independently verify, so the area geometry is clearly a candidate for removal.

I also checked a few areas where inside the area is only a single settlement and that one has also a place node with a name.

Unfortunately only a fraction of these way areas had names which correlated with the existing place names. If not assuming that all place nodes in OSM are from some bad sources, I would rank their quality higher than the names on fictional area boundaries.

So in this case, the last potentially worthy bit of detail got also void.

Based on this I would vote for a complete delete.
If someone could verify that the name tags are actually valid and just a “Ban” is missing in the front we could fix that. But not certain it is worth the risk/effort.

Stephankn, could you give some examples of how the names conflict? From what I’ve seen they didn’t seem doubtful.

The areas appear to be manual traces of neighbourhoods rather than official boundaries, so retaining only the names if possible seems a fair suggestion.

Paul, Elijah told me the areas are meant to include the settlement as well as any surrounding area that could be a breeding ground for malaria-carrying mosquitoes. I don’t think they serve a useful purpose for normal users of OSM data.

I vote to delete them. I say forget about trying to salvage a tiny bit of useful information from them. It’s too much work for a very small return.

@Paul: sorry, I didn’t copy the references to the not matching ones. Can’t remember which area I checked. So maybe the accuracy of the names depends heavily on the region/mapper drawing it. I opened the filter above and thought I might come up quickly with bad examples, but this time only spotted names which had been duplicates (the areas typically lack “Ban”).

Still the geometry is nothing I would keep, as it covers way more than the residential landuse of the villages. Maybe you could take-over the name as place name if something hints that it is correct. Neighboring villages with the correct name from same mapper might be a good indicator. Or Wat/School with the same name. Still a lot of cleanup work.

If you have the capacity to review some of the areas and copy over the useful details into a place tag, it would be highly appreciated. If you do 10-15 each day it is done in a month. The total number of potentially added place nodes is then in the area of 300. A spatial query could give more details. Maybe checking that out later.

Here a query which selects the areas in questions where there are already tagged places, Currently 71 polygons matched, some cover multiple place nodes (77 places).

In theory these areas are duplicates and could removed right away. They might serve as an indicator how well they match. So untagged place nodes could inherit the place names.

As I want to avoid missing to document findings like last time, here the double-nodes:
way 691673750 name=นุโพ, node 1440277185 name=แม่จัน, node 7099668518 name=Nu Po Refugee Camp
way 691890323 name=เปลิ่งเคลิ่ง, node 1444460643 name=บ้านเปิ่งเคลิ่ง, node 5040884680 name=บ้านเปิงเคลิ่ง
way 691863974 name=กลัอทอ, node 1858019059 name=แม่จัน, node 5040726390 name=บ้านกล้อทอ
way 691851486 name=อุ้มผาง, node 795764576 name=บ้านโชตา, node 795761644 name=อุ้มผาง
way 691849701 name=วังผา, node 5035884677 name=บ้านวังผา, node 795672667 name=บ้านไม้ริมเหม่ย
way 691844773 name=หนองหลวง, node 3997932144 name=บ้านหนองหลวง, node 2103966229 name=บ้านหนองหลวง