In the West Midlands we’ve been busy with rural mapping. To those of you surprised that there’s any countryside in the sprawling industrial conurbation that spreads from Wolverhampton through Birmingham and Solihull into Coventry, we need only quote the motto of Solihull which is Urbs in Rure (Town in Country). You will see that there is a large amount of green belt land that has been jealously guarded by planning authorities. The green belt land provides much needed recreational access for the surrounding population. It will come under increasing pressure from the new relaxations on planning restrictions.

In a COUNTRYSIDE TRIAL AREA to the eastern end of Solihull and the Western end of Coventry bounded by A45 / Birmingham Road Meriden / Berkswell Road / Meriden Road / Coventry Road / Broad Lane / Eastern Green estates / Allesley Green estates, mappa-mercia mappers have collaborated to produce a rural map to the most detailed level currently available within OSM.

Guidelines

The guidelines which we evolved by discussion, some concentrating on mapping landuse and others on mapping footpaths, are as follows:

1. All field boundaries are shown using the predominant feature defining it to someone in the field. The most useful are: 
– barrier=hedge
– barrier=fence
– waterway=stream (NB: if a stream also has hedge one side and fence the other we only show more than one if it is necessary to make situation clearer).
– highway=service (or higher – only show barrier as well if necessary)
– boundary between landuse=residential and landuse=farmland (NB – these are not so easy to see and therefore adding a barrier as well is desirable, or for “MapQuest Open” type rendering, essential.

2.   Adding the main buildings within a residential area is desirable in order to give as god idea what the area looks like. We  have tried to add all houses, and usually all similar sized or larger buildings to all small areas of landuse=residential. In some cases we have added all buildings and the dividing hedge/fence between plots, but regard this as optional and maybe overkill/over use of memory/rendering-CPU-time/etc. In larger residential areas such as Meriden and Berkswell we have not added any buildings to contrast village with hamlet/mansions/etc., but would welcome other local editors to take on such tasks over time. In larger areas still, it is obvious by adding all the residential roads that the grey area is residential. Buildings for farm use are default labelled as building=agricultural, unless the specific use is known, e.g building=glasshouse

3.   We have  tried to give a landuse to all light grey areas other than the verges of roads and some tracks and footpaths. Sometimes this needed lateral thinking to define an area by its major use to avoid leaving light grey bits of scrubland, etc.

4.  Being surrounded by large residential areas where horse riding is popular amongst the residents there is  a large amount of farmland devoted to grazing and exercising horses: these we have differentiated by using landuse=meadow. It is often hard to get this right, for example the meadows north of Pickford Grange Farm are usually full of horses, but checking with Google StreetView shows cattle in the eastern 2 or 3  so we’ve left them meadow). There are other areas which are probably anomalies but regular ground surveying and more discussion will eventually clear these up.

5. Landuse=forest has only been used where it is obvious by the regular pattern of the trees that it  is a planted and managed area of trees. All other wooded areas are tagged natural=wood. Currently we are not sure that we have been totally consistent here. There can also be ambiguity between a line of trees, a thin wood and a thick hedge,especially along watercourses. Much here is a matter of styles between individual mappers.

Issues

A. The brown colouring of the farmland we hope one day soon be made lighter/brighter. The current brown is not, to us, aesthetically pleasing, and  makes seeing highway=footway paths hard to see. The residential grey has almost as bad an effect. 
It may be a good idea to render footways somewhat wider (especially at larger scales), and/or increase the width of lightening the background which works so well with woods and forests.

B.  We have been adding a lot more of the stiles and kissing gates. Unfortunately the latter do not get rendered whereas gates, bollards and stiles do. Kissing gates locations are useful to know if you need to navigate with persons who can walk adequately but have mobility difficulty in crossing stiles or you have a very large dog which you can’t lift over stiles bit you can get though kissing gates. We can’t rely on the absence of stiles on the map to infer a walkable  route in these circumstances (incomplete surveying), but the explicit rendering of kissing gates would be a great asset.

Boundaries, be they land ownership or administrative boundaries, form an essential part of geo-data. Was it therefore a success for the open data campaign when Ordnance Survey released the Boundary-Line product, containing all electoral and administrative boundaries, under the OS OpenData Licence? Not Quite.

The problem is that boundary is the “Boundary-Line” data are not boundaries! There are two issues, and these apply to all geospatial data released as independent vector layers; scale and context.

Scale

Large scale maps (such as walking maps), have a higher resolution than medium, or small scale maps (such as country of global maps). The larger the scale, the higher the resolution and therefore the more detail shown. As boundaries twist and turn following streams, rivers and hedges, it is important to use a large scale/high resolution. The Ordnance Survey provide boundary data in two products; the free “Boundary-Line” product, and the non-free “OS MasrterMap” product. According to Ordnance Surveys own admission:

Boundary-Line is captured against a lower resolution mapping backdrop and the boundaries are captured to represent the data at a nominal 1:10000 viewing scale. The process of generalising the data may have caused some features to be moved from their true ground position for the purpose of map clarity.

So scale, or to give it it’s proper name, generalisation, is the first problem we should be aware of before using Boundary-Line data in OpenStreetMap.

Context

Lets see what Ordance Survey have to say about their other boundary product – OS MasterMap:

OSMM Topography Layer currently holds the definitive and more accurate boundary information as the boundaries are mered (aligned to) real-world features on the ground. Captured at mapping scales of 1:1250, 1:2500 (for urban areas) and 1:10k (for rural).

Okay, we can see that MasterMap uses a larger scale / higher resolution which helps provide more accurate data, but what is that reference to “mereing” (alignment). Mereing, is the process of establishing a boundary relative to ground features present at the time of a survey (source: ESRI). As such any boundary line extracted from a map of any resolution cannot therefore be used on its own to describe a boundary – once you extract the boundary, you loose the context provided by the rest of the map!
Lets look at an example.

In the example above we see the boundary as a black dashed line. Along the line are some descriptions in pink. The first couple are relatively easy to decipher – “Co Const Met Dist & CP Bdy” is shorthand for County Constituency, Metropolitan District and Civil Parish boundary. Similarly “Boro Const Bdy” tells us it is a Borough Constituency boundary. As we follow the boundary we reach more abstract shorthand:

• 1.22m RH – RH stands for Root of Hedge, and 1.22m equates to exactly 4ft. The boundary is therefore 4ft away from the stem line of the hedge. This unusual convention comes from the presumption that landowners planted hedgerows slightly in from the edge of their land so as to not encroach on their neighbours plots. The distance differs between 3ft and 5ft depending on which parish the land lies within.
• Def – This is a “defaced boundary”, meaning that the original feature that the boundary was aligned to no longer exists. This was probably a hedge that was removed when the houses were built. Had the hedge roots not been removed then the boundary would be marked as “Tk H” implying track of hedge.
• Und – An “undefined boundary” is one where there was no real-world feature to align the boundary to when it was surveyed.

For a list of abbreviations click here.

Conclusion

Any vector extract of boundary data will be problematic due to generalisation (scale) and the loss of information about mereing (loss of context). If you are really keen on determining the exact boundary you may want to read more about boundary presumptions or go and dig out the original surveyors notes (Perambulation Cards) at the National Archives.

Discussion

So for OpenStreetMap should we be linking boundary ways to streams and hedges on the acknowledgement that Boundary-Line data is not perfect and we are not able to provide any certainty as to which side of the hedge the boundary actually resides on? Please leave your thoughts as comments below.