Street Lights: Local Government Open Data 3

Another data set released by my local council contains the locations of all the street lights owned (or operated) by the council. Now I have no intention (or any desire) to import this data into OpenStreetMap. Is it any use in the open data context?

There are two obvious questions that can be asked using street light data:

• How well mapped are all public highways (i.e., including footways and cycleways) in the city?
• Are the highways on OSM accurately mapped?

Street lights are spaced at specific intervals according to the type of highway, it's width and the nature of the light source. This means that their accurate placement is determined by the council beforehand, all location data is to the nearest metre. Since most lights are situated immediately adjacent to the road carriageway they will be a short distance from the road centre line (say under 10 metres).

It is trivial to buffer the network of highways from OSM (via a Geofabrik download) and then use geometry operators to find all lights within the highway buffer envelope, and all outside the envelope. Using a very simple constant buffer of 10 metres I obtained the following result:

Street lights which are not close to OSM roads (best viewed large).

The most immediately noticeable thing on this map is that some areas have a much denser pattern of 'misaligned' lamp posts. In particular there are several areas in the far north of the city: these are all social housing developments dating mainly from the 1970s and 1980s. Although there are relatively few missing streets, there are many (named) footways in these developments which are missing. In addition many of these areas have not been received a field survey so this type of feature is missing. Even in other similar areas of the city (to the S and E of the City Centre) which have been surveyed there are still many missing paths, at least in part because the path network is very dense and difficult to survey. In the main this information corroborates what I already know about which areas of the city require more detailed surveys.

part of Rise Park Estate, Nottingham showing missing footways
on the basis of street lights not aligned with OSM data (yellow),
remaining lights are shown in maroon

A second. aspect is accuracy. Many streets have 'misaligned' street lights on their S side. This is most obvious with streets which are oriented E-W. I have a pretty good idea that this is because the streets are indeed misaligned, most probably because they were traced from Yahoo imagery long ago and the alignment has not been corrected since. (And for good reasons, realigning data and keeping the topological relationships right requires a lot of care).

Potential mis-alignment of OSM highways in Wollaton and Bilborough areas of Nottingham
(coloured blobs are street lights as above)

Thirdly, we can find missing highways. The example below is particularly clear because it shows a cycleway through a park.

Missing foopath/cycleway at Sandy Banks Local Nature Reserve, Nottingham
(the path has been partly mapped since from Bing imagery,
a survey is required to check if it can be used by cycles )

One last point, is that one can also find areas of land which are private (some are well known, such as The Park shown below, which is all lit by gas), but I've noticed one or two new developments with no council owned street lights. (I know the dataset is up-to-date because it includes the positions of new lights on our road installed in the Autumn).

The Park, Nottingham, showing absence of council-owned lighting.

At some stage I'll work a bit harder to see if I can quantify any misalignment factors. I'd also like to see how feasible it is to re-create the road network just from the position of street lights!

Once again the principle message is that this is another data source which can be used to validate, check and improve OSM.  It's particular value is in identifying where we have failed to map footways and cycleways: I am not aware of any other (open or public) dataset which enables such a comparison.