Our shitty map

We have threads about the bad quality of Facebook imports, and also about Grab’s work.
Just recently, I changed some “residentials” created by Facebook to streams.

But let’s face reality: there are many more bad things in our map than Facebook and Grab can achieve alone.
I’d like to give here a few examples as I work on my collected GPS traces of last December.

Today I want to show this example:
I cycled along way https://www.openstreetmap.org/way/106995052
It was freshly asphalted just recently:

But let’s look forward:

There’s the end of the road. A car parking on the left, there is a hut.
Nothing follows behind it. No track, but one is shown on the map

(Larger version here)
Perhaps the (small) irrigation canal was mistaken for a track.
Well, I returned then…

I think if you look at Google maps and Streetview there may be an explanation for this. On Bing a track of some sort is clearly visible, at least for a short distance.

What boggles my mind is the bureaucratic thinking behind re-sealing such a dead-end road. Oh wait, the man living on the left must have important connections… :roll_eyes:

Connections between non-connected roads are also annoying. Again an example from Loei:
Residential https://www.openstreetmap.org/way/106751998/history#map=18/17.48445/101.72679 was connected with https://www.openstreetmap.org/way/662613675#map=18/17.48555/101.72676 (and depending on the zoom level, you might still see them connected).
But coming from north (near the petrol station with Cafe Amazon), you will encounter some kind of gate preventing you from getting to Baiboon Place hotel.
(Well, true, the worst thing with Loei province is the imagery: typically off by several meters, the offset changes every few meters…)

They thought the “track” at its end is part of that road, but then finished their work when they detected that it did not exist :slight_smile:

I see. The job is completed and the entire budget is spent.

I left Ayutthaya in the morning, and in the late afternoon a short rainfall caught me in Kaeng Khoi. Fortunately, the rain was very short, so I decided to travel further towards north east. Shortly after the end of the village along road 3224, there is a resort (I did not add a waypoint on my GPS).
According to our map, there’s a guest house not too far away:
When I arrived near that place, I found the road closed by a gate (the building belongs to Siam Cement Group). I asked the security guy about the guest house. No, it is not there. But there’s one a few kilometers away, and he gave rough directions. I searched for it in the next village, but it was a little farther away. Eventually I arrived at
(almost back on road #2) which did not exist on our map (I estimate that far more than 90% of “resorts” are still missing). Then the rain started again.

The private roads of the quarry / cement factory nearby were imported by Facebook, and not marked as private roads.

On the morning of Dec 13, I left the village on the Mekong river towards the major road 211 on an “unclassified” road
Suddenly, the concrete pavement ended.

Well, the major road is not far away, so I can try that track.
Shortly after the bend to the left, things changed again:

A terrible rather overgrown path (even more so in the parts not visible on this photo), which is used by people on moped sometimes.
I tried it, as the major road was very near. And fortunately, it connected to it.
Oh, by the way, Google Maps does not know this great “unclassified”.

I took a look at the hotel, and perhaps Im the guilty one. Loei town was very barren 8 years ago, and the imagery not good so it possible I “saw” a connection. I stayed there at the Muanmanee last night which prompted me to look at the road outside - it now runs right through the middle of the building.
Yup, someone has helpfully moved the road to the “correct” position, but did not move anything else nearby. But he didn’t stop there, he moved the whole of Hwy 201 across too, to where he thought it should be. Of course in his hurry to help, the junction with the 21 now looks like a godawful mess. The cyan GPS traces cleary show the correct location.

Then we have a random selection of other roads that were moved, and some not. I can see from the history who it was, but I wont name. He also seemed quite keen on inputting loads of nodes simply tagged as shop=yes. Yes that’s a great help !
And then to compound the problem, Facebook came along and added a load more to the poorly aligned imagery, exacerbating the problem.

When I get 2 days free, I’ll correct the whole of the City, and title the changeset “Dont f**k with Loei any more”

One final question to you - I see you are tagging some roads as dual_carraige=yes. This even includes the link roads. I can’t see anything in the Wiki about it. Can you explain its use, so we can use it too, assuming it carries some relevance.

Still, I would not call our map “shitty”. Outside the cities it beats Google and Esri hands down, unless they have significantly improved lately.

I doubt that it can already be fixed. Most GPS traces are for major roads, hardly any traces for minor roads. And the alignment of the imagery is too variable, the offset changes too often (also when looking at the traces by others, I know that my Garmin Oregon produces quite shitty traces every now and then).

E.g. near Muanmanee (I stayed there 2 nights btw) the offset of DigiGlobe Premium is bad, while it is great near the Northern end of 2138. In the West of the town near temple https://www.openstreetmap.org/node/6205240641 it is bad also.
You cannot just select many roads and drag them to the “correct” position, when the offset at the start of the road is different from the offset at its end. I did not dare to fix the town.

I added a long cycleway next to 201 which runs rather parallel to the road according to my GPS trace - that’s now quite away from the main road on the map, and sometimes intersects other roads at likely wrong places - not good. But without GPS traces for the minor roads, the offset at those junctions cannot be estimated.
Now we have the problem of some roads aligning with some imagery, others with GPS traces, others with neither…

dual_carriage is an inofficial tag. I use it here to detect dual carriage trunk-like highways and render them as trunks. A few years ago, we had a long discussion on tagging of highways, but … So I use it now even more often, and there is no need to interfere with tagging by digit count which is preferred by too many people here.

No, our map lacks almost all Points of Interest. There are e.g. no “ressorts” in Khemmarat, are there? Well, in a couple of days, I’ll add about a dozen there. Our map showed me resorts in Laos instead. Also a question: are there so many more resorts (also restaurants etc) in Laos than in Thailand? Or how does it come they are so much better mapped in Laos?

Ok, I was referring to routing.

Hm… In Dec 2017, I took a taxi from air port to Petchraburi. The taxi driver asked his phone “OK Google 2N guest house Phetchraburi”, and thus found the way. On my Garmin, I had a map built for bicycle, and only after having crossed Bangkok it could calculate a route. The same route as did Google.
In Dec 2018, it was Ayutthaya. Google could calculate a way to the guest house, but actually failed in the town because of the oneway roads. OSM had them correctly. But the guest house was not on the map (well, I extracted its position from Google Earth, so routing did work).
I think that finding the target on the map is part of a proper routing application.

Let’s look at a different place:

In Google, that’s just some kind of “road”. See https://photos.app.goo.gl/Qb63JzqTuzfEXk5W7
In OSM, that’s a track (a Facebook import). But the track was not split at the position where it becomes really bad.

Next place. In Nong Khai province, near the Mekong.
Road https://www.openstreetmap.org/way/287851890/history is completely inside Nong Khai province, no centimeter is in Udon Thani province. None the less, it’s ref number was “อด.5028”, i.e. containing the abbreviation for Udon Thani.
The situation on ground is not great either: at its beginning, there is a DRR sign showing “นค.5028” (i.e. the abbreviation for Nong Khai province), followed by a DRR sign with “นค.3028”. And the milestones used of them, I don’t remember which one.
Nowhere the abbreviation of Udon Thani was used.

Road https://www.openstreetmap.org/way/287825170/history is in a school. Both ends were connected to the minor road nearby. But actually, only its eastern end is connected - at the other side, there is a wall.
Again a non-existing connection, not caused by FB or Grab.

At least, there was the “reference number” of “submerged” on the map:

I came from south, and at the junction there was a big sign board (actually inmidst the old road) reading “road closed” (in Thai only, of course). I had time and have fun looking around in the country site so I cycled along.
Yes, the road is submerged.

Tertiary อด.4028 turns left at that junction. That road was on our map, but as an unclassifed only. There were further mistakes to be corrected with อด.4028 later on.

OK, I have to hold my hand up to that one. I can’t read Thai, I don’t have a Thai keyboard … so its a cut and paste for me to get the Thai province prefix. Ironically, yet both Garmin and Google have the adjoining 4028 down with the Udon Thani Prefix, so I guess Im claiming mitigating circumstances :slight_smile:

Is there anyone clever enough to write a program to check we have the correct prefix for every provincial highway, according to the province it is in ?

Yes, that can cause confusion. It has an Udon Thani prefix, but it runs far south into Changwat Udon Thani, so it can have the “foreign” prefix.
And that’s the thing which makes a tool for checking the prefix hard. Just remember a thread from last year: https://forum.openstreetmap.org/viewtopic.php?id=60932 where I reported a Kanchanaburi number in Phetburi province.

Regarding typing the prefix: I normally don’t type it, but select a number with the correct prefix from the list in JOSM, and then change the digits. But I know (most characters of) the Thai alphabet, so I can see such discrepancies (but they are still hard to see - such a number has no meaning, so I may mis-read it also).

And an important point: I do not want to blame any mapper for the issues I show here. My point is that also we (i.e. including me) committed quite a lot of mistakes in the past, not only Facebook and Grab (and I don’t say that their quality is good…). It is just unfair to them that some people say “before FB we had a great map, and now it’s such bad we cannot use it anymore”.

I second that comment about the fact that we make errors as well. Every mapper makes the sorts of errors we moan about when taking FB to task for theirs. OSM employs a somewhat chaotic tagging system that takes quite a bit of experience to learn.

Also, an aside: As I read through these posts, and those in other threads, I am amazed that more errors don’t show up. Some of you actually make notes and add them to waypoints on your Garmin devices and then transfer that data to OSM after the ride. That is such a tedious and error-prone process! You can make your life a lot easier if you buy a cheap camera and take photographs of your POIs, street names and route numbers. Then, using JOSM’s photomapping plugin, add a time correction to the EXIF data on each photo to place it spatially on the map. Even easier is to use a smartphone camera with location services turned on. Using such a workflow, you will have a photo of each POI that has its location and other salient information right on your computer screen.

As for prefixes, I use Autohotkey (a keystroke macro program) to add Thai prefixes for the various provinces when I’m working with route numbers. Tom Layo turned me on to that s/w a few years ago and I use it constantly. For example, when I want to add the prefix for Chiang Mai to a route ref, I simply type cm followed by two dots (so it isn’t confused with the legitimate abbreviation for centimeter) and it types ชม. I have such “translations” for all the Thai provinces always available. It’s an awesome productivity enhancer. I also use it to expand common words like school (I type rr followed by two dots, which gives me โรงเรียน) and to automatically fix common spelling errors, sort of like a custom spell checker. I can’t tell you how much work this little (free) addon saves me.

My two-cents (two-baht?).


Thought of that quite often. But beyond being a cheap camera, it must be easily accessible while I am riding my bike. I don’t want to stop, take the camera out of the bags, take a photo, etc… Questions about such mapping gear get asked in other parts of the forum every now and then. Now good answer available yet.

Just started to work on the traces from Buengkan province. I wanted to add a reference number for a DRR road. Opened the tagging dialog of JOSM, typed ref in the key field, and opened the list of the value field. All suggestions with a prefix started with the Nong Khai prefix. I downloaded further data, a larger area even than required for the 100 km trace I am working on, and no road had the desired prefix. I copied it from the node of Buengkan town then. I’ll change the prefix of those roads which do not cross changwat borders.

The “Autohotkey” utility wouldn’t help here - I have to get บก from somewhere first.

I eventually downloaded the area of Buengkan changwat, and changed the prefix. With the exception of one road leading into Sakhon Nakhon changwat, all had Nong Khai or even Udon Thani prefixes.
Though the prefix is now fixed, the numbers are likely not so correct: too many road numbers start with digit “7”.