OpenStreetMap

Mapping addresses helped find a nine year old mistake in OpenStreetMap (also in Apple Maps, HERE, and TomTom)

Posted by nickjohnston on 15 June 2020 in English (English)

The need for addresses

One of the things that often comes up in discussions about OpenStreetMap is poor address coverage in some places.

Rightly or wrongly, to many people a map is now no longer just a 2D representation of our environment. Searching and navigation/routing, despite being separate services, are viewed as part of a map, and almost as important as the visual map itself. We need good address coverage to improve geocoding.

Adding addresses really improves the map, and that’s what keeps me going with it. When I explained what I was doing to one person who approached me, he said “that’s a good idea mate, Google Maps shows my house all the way down the end of the road”. I also saw a food delivery driver pull up outside a house and ask “is it called Gate House?”. (On OpenStreetMap it is ☺)

Chancel Way or Chancel Park?

Recently I collected addresses on Chancel Way in Charlton Kings, Cheltenham. When adding them, I noticed something that didn’t seem right.

The road was split into Chancel Way and Chancel Park. At first glance, this seems fine. Having a common prefix like this is common in UK residential street names. (For example, Croft Road, Croft Gardens, Croft Drive, and Croft Avenue are nearby.)

The house numbering suggested that Chancel Way and Chancel Park were actually a single street. Following the street west, the first number on the left (south) side of Chancel Park was 19, and the next number was 21. This fits with the final odd number on Chancel Way (17). Even numbers on the right (north) side of the street match too.

More research

There are no results for Chancel Park on the Royal Mail postcode finder, yet it appears on Apple Maps (via DuckDuckGo):

Apple Maps showing Chancel Park

It also appears on TomTom (no surprise as Apple uses TomTom data in places):

TomTom showing Chancel Park

The mysterious Chancel Park also appears on Bing Maps:

Bing Maps showing Chancel Park

… and HERE.

Much as it pains me to admit, Chancel Park is not shown on Google Maps; the whole street is shown as Chancel Way.

I even walked back and checked for any sign of Chancel Park on the ground. I found none. Chancel Park does not exist.

How did the mistake happen?

Chancel Park was added in changeset 7794450 on 2011-04-07, over nine years ago. The source:name tag was set to OS_OpenData_Locator. To my surprise, OS OpenData Street View shows Chancel Park:

Chancel Park

OS MasterMap correctly shows the whole road as Chancel Way though.

It’s likely that TomTom (who supplies Apple) and HERE trusted OS OpenData Street View, and ended up with the mistake that way.

Closing thoughts

I discovered this mistake (which I’ve now fixed) through address mapping. If I’d just looked at the area on OpenStreetMap, the problem wouldn’t be obvious. Even a normal ground survey wouldn’t pick this up unless you were specifically comparing streets on the map with streets on the ground.

While out collecting addresses I’ve also found other useful things to map including:

  • Tiny missing paths and shortcuts (like this one)
  • Missing turn restrictions
  • Cycle lanes

Looking at other maps for preparing this entry has made me realise how plain they are and how little information they convey compared to OpenStreetMap. Apple Maps has no buildings in this area, and doesn’t show River Chelt either.

Most striking though is how little information other maps show for walking and cycling. There’s a path linking St Judes Walk to a walking and cycle path to the west. This is a pleasant route through Charlton Park and Cox’s Meadow, and then on to the hospital, lido, Sandford Park, and the town centre. This level of detail is a huge advantage of OpenStreetMap.

Location: Moorend, Charlton Kings, Cheltenham, Gloucestershire, South West England, England, GL53 0ND, United Kingdom

Comment from Sanderd17 on 16 June 2020 at 13:55

Regarding the Chancel Park/Way. I wonder if it’s some copyright easter egg that got included into OSM.

It’s an ideal easter egg, as the original name can still be found. And very few people will notice that street has two names.

Perhaps the data was once meant to be under a closed license, but later released, and the easter egg got in that way?

Comment from SK53 on 16 June 2020 at 18:18

@Sanderd17 : Ordnance Survey are on record as to not having easter eggs in their map & data products. They have used ‘signatures’ in legal cases (notably vs. AA) based on a collection of features which it would have been unlikely to duplicate by chance.

The original OS Locator & associated products had a significant number of errors: some minor, but others with names completely wrong. Some examples:

  • Forest Green Road, Ockley. I surveyed this around 6 years ago and filed a bug report to the OS. Consequently Google, Bing & Here all have it corrected as well as OSM.
  • Smythson Drive. This was originally named Smithson Drive, but given that a notable Elizabethan architect, Robert Smythson, not only designed the local mansion, Wollaton Hall (aka Wayne Manor), but is buried in the church,it was a good guess that this was a spelling error easily checked by reviewing one of my survey photos.
  • Gregory Street. One of many roads where the change of name was inaccurate, see my detailed account of such errors. Incidentally the evidence I used rather goes against the statement by a prominent British politician about old maps.
  • Turing Gate. My favourite, because a bit of contextual knowledge strongly suggested that this road was named after Alan Turing and not called Turning Way (not that the OS social media team got this). It’s in Bletchley close to Bletchley Park; other roads have names related to the cryptographic work there during WW2 (Colossus, Ultra, Hinsley, Enigma); and Turning Gate is a daft name.

Virtually all these points just reinforce what nickjohnston says in his post

Comment from GinaroZ on 16 June 2020 at 22:00

While out collecting addresses I’ve also found other useful things to map including:

Another thing that is good to map after surveying is speed limits - I notice none of the residential streets in the area you surveyed have a maxspeed tag :)

Comment from Sanderd17 on 17 June 2020 at 10:56

How does Ordonance Survey gather the data? If it’s really by surveying, maybe it had a wrong sign somewhere in the past?

My own street had a wrong sign for a long time. The correct name is “Vyvestraat”, but for many years it said “Vijvestraat” on one of the signs. For historians, ‘ij’ used to be a single glyph in Dutch, but after the invention of the printing press, it was costly to produce that glyph, and generally was printed ‘ij’ in the Netherlands, but ‘y’ in Flanders. Later on, Flemish spelling evolved to follow the Netherlands, but old names still have a lot of ‘y’ characters in them.

I once even found a street with 3 different spellings (also ij vs y related): https://www.openstreetmap.org/way/132846607

Comment from gecho111 on 21 June 2020 at 15:40

I found a several local very old misnamed streets using http://osmose.openstreetmap.fr/

A few years ago I did an address import which had the official street names. Recently when I decided to tackle Osmose errors (with my name on them) there were many for (“addr:street” not matching a street name around).

Comment from trial on 24 June 2020 at 20:50

@gecho111, that’s why In France (not in Germany ;-)) people like to use associatedStreet: points of adresses just need an addr:housenumber, the street name is in the associatedStreet and the street members of the relation. Still some duplication but much less. Add the possibility for multilingual names: adding on each single POA is just a nightmare. When streets get renamed, it’s easier too.

Login to leave a comment