Hackney: “Complete” on OpenStreetMap

ITO‘s OSM Analysis table is updated daily, showing the number of roads in each district/borough in the UK that are in the OS Locator* dataset that are missing in OpenStreetMap. There is an accompanying choropleth map (you need to login to seet it) showing coverage across the UK. Currently southern England and Scotland’s central belt are looking pretty good, while Wales and parts of northern England still have quite some way to go.

As well as the map and summary table, ITO produce a map showing the approximate location of each missing (or misspelled) road, as a rectangular bounding box. This makes finding the missing roads quite easy. Groundwork is needed to check signposts and confirm the names. Often, discrepancies arise simply from OpenStreetMap not having apostrophes for the street name, and the Ordnance Survey having them, or vice versa. The signposts normally provide the definitive answer, but in quite a few cases, the sign at one end of the street will have an apostrophe and the other end will be missing it – or the names attached to houses will differ from the nearby street signs.

Hackney was around 94% complete a few weeks ago, with around 80 errors flagged up. It’s a small borough (in size, if not population) so I reckoned it was possible to bike around all 80 locations in a day – with a suitable route that would hopefully be a good answer to the Travelling Salesman Problem.

In the end, I managed it in a couple of afternoons. I found quite quickly that in most cases, the street was already in OpenStreetMap but just misnamed or unnamed. In around 5-10% of cases, the Ordnance Survey was wrong, and street signs on the ground suggested either that OSM was correct, or that both sources was wrong. As the OS was in general right, I only stopped to note where this wasn’t the case – so, I was able to cover a lot of ground quite quickly. I still ended up having to cycle nearly 100km within a borough that is roughly 4-5km across.

Perhaps the most interesting part of the exercise is it allowed me to visit all parts of the borough, including places I had never been to before despite having lived there for several years. I discovered that Stoke Newington Church Road is indeed massively gentrified out of all proportion to the surrounding area. I found unexpected gems like The Mothers Square (no apostrophe) and that Hackney Downs is quite a pleasant park. I also found the huge Pembury Estate. This was where one of the Ordnance Survey’s mistakes was – the streets were named in the wrong order. Perhaps the original surveyor didn’t like to hang around.

Of course no map can ever be 100% “complete”. Even with a perfect match to OS Locator, the latter may be out of date, or be missing roads due to missed records. The map may be “complete” in terms of roads but other detail still needs to be added.

Anyway Hackney is now 100% complete with respect to OS Locator, making it the 12th such district in the UK. Another 400-odd to go…

* OS Locator is one of the products in Ordnance Survey’s OpenData release.

6 replies on “Hackney: “Complete” on OpenStreetMap”

The incomplete condition of Open Street Map allows us to remember also the human condition of the geography. I think that the walk around the place (Hackney) also is important in the work of a geographer. I liked much the brief but precise description of the zone as scientist (gentrification of the place) and flâneur (the unexpected gems).

I think I understand what you’ve said; thanks for the reply anyway.

One particular example is Ffordd Abergele in the county of Conwy. OSM Analysis is looking for Abergele Road, even though I’ve put in name=Ffordd Abergele and name:en=Abergele Road.

I thought that OSM Analysis got a list of street names from the local authority or OS, and was looking for those matches based on the name tag.

Huw – it all depends on what’s on the street signs. Generally, road names don’t get added in to OpenStreetMap except by people visiting the street concerned, so if the street has the Welsh spelling and the has the English version, then the Welsh version will remain and a not:name (or possibly name:en) tag added to acknowledge that the OS discrepancy has been checked.

If this is not the case, feel free to change the name to correspond to the street signs, and add in a not:name or name:en additional tag with the OS name in it.

A problem more specific to Wales is the choice between Welsh and English roadsigns.

A road locally known as say Ffordd Elwy, may be registered as Elwy Road (that what it actually means) by the OS. Although not incorrect, it’s a bit cheeky making an official survey only to name something differently to local use.

This of course means it will not be picked up by OSM Analysis.

Mark – I agree completely, the OS is not definitive, however it is still a good way target specific areas of the OpenStreetMap map in terms of improving its accuracy. This is also why each street was visited – in most cases the OS was better, sometimes OSM was better, and sometimes both sources were wrong. What is on the ground takes precedence other OSM or OS, or even over council WebGISes, unless there is clearly an obvious mistake.

The only problem with this is that the OS are not the definitive source of road names – it’s the local authority, they are the only bodies who have a legal right to name roads but getting the different data sets to be the same has always been a problem – as you’ve found out, even the local authorities can’t get it right on the same street!!!!!

Leave a Reply

Your email address will not be published. Required fields are marked *