Wednesday, 22 July 2015

OSM Retail Survey: Part-3

It is quite difficult to identify areas of incomplete retail mapping within a large conurbation.

Initially I thought investigation of areas marked as “landuse=retail” showed promise, but they proved disappointing in practice. The meaning of this tag has been interpreted in different ways, and used inconsistently. Almost half of retail areas in OSM contain no retail properties, so they can provide some indication of gaps. However, some of them are very small, and two-thirds of retail properties already in OSM lie outside a marked retail area, so in general checking for data on shops within retail areas is unlikely to be efficient.

  • A number of fairly large settlements don't have any retail area described in OSM (including Weymouth, Wellingborough, Grantham, and Newark-on-Trent). 

The reasons vary.

  • In most of these the main retail area hasn't been marked as such, even though other types of urban landuse have been applied in other parts of the town. 
  • In some cases the whole town is marked as “landuse=residential”, 
  • In others an area that seems to be predominantly retail space has been described as “landuse=commercial”. 

Overall, there are too many exception cases to make productive use of retail landuse data.

On a very traditional high street, individual shop frontages tend to be relatively narrow, and relatively consistent in length. In theory, if we could compare the density of shops in OSM with what we expect, then we should be able to get a sense of where the database looks thin. However, this can only give a broad indication.

  • A well-documented street with a few large shops will still look empty in comparison to a street where the data only includes a small proportion of numerous small shops. 
  • There are practical difficulties in calculating density consistently when nearby shops may lie on opposite sides of a street, or in adjoining, and neighbouring streets. 

These challenges can be overcome to a degree, but my attempts have involved some intensive computation. In practice a simple heat map seems to be equally useful for flagging up suburban areas that already have fairly high levels of retail content. In conjunction with local knowledge of where retail outlets are clustered this could be sufficient to identify some larger suburban shopping areas that need attention.

This is an example from Newcastle. Data on retail is dense in the centre of the city, the quayside, and along Gosforth High Street. Coverage of retail barely shows up in areas such as Jesmond, and along Westgate Road. Those with local knowledge might be able to use this kind of feedback to identify areas that are worth further investigation.

Finally, while the overall level of retail coverage in England is 27% of retail premises, there are variations in the extent to which different types of shop are recorded. Analysis of the mix can be useful at a national level, but it can also be informative at a local level.

First, though, we will look at the mix at a national level.

No comments: