Housing Terraces in Wales : a minor addressing conundrum

Two terraces of housing off the Holyhead Road (former A5) in Llanfair PG.
Penucheldre and (to the right Britannia) Terrrace. A typical type of housing throughout Wales, they provoke some addressing conundrums.

In May 2015 I attended the funeral of a relative in the Anglesey village of Llanfairpwllgywngyll (usually referred to as Llanfairpwll or Llanfair PG, but perhaps better known for having an extremely silly long form of the name). This is the ancestral village of my maternal great-grandfather: the funeral marked the end of around 150 years of members of the family living in the same house that he grew up in.

I know the house well. As a small child I visited frequently. It was only a short excursion by bus from Bangor, where we lived. My mother knew it much better: in the early part of WWII she and her grandmother lived here with my great-grandfather's sister. Even when I was small the village was not very large. When my mother was a girl there was much open land between the old part of the village (Pentre Uchaf) and the newer part (Pentre Isaf) next to the railway and main road to Holyhead. By the late 1970s the village had grown immeasurably with lots of overspill suburban housing for Bangor.

The big change between 1939 and the end of the '70s was that streets started to be named and houses numbered along the street. Prior to that building development had been piecemeal: most usually a mix of individual houses and most typical of many parts of Wales: named terraces. By this I mean short terraces of houses where the terrace rather than the closest street provides the name used in the address. Elsewhere there are plenty of terraced houses where individual terraces have names (often shown on a carved stone set into the brickwork of the terrace), but the numbering of houses solely relates to the street.

 Terraces in Llanfair PG

Looking towards the main road (Ffordd Caergybi or Holyhead Road). Two terraces are visible: one on the L angled at around 60 degrees to the street, and the one on the right, Williams Terrace (I think) facing onto the street. There is another terrace behind Williams Terrace, and two more off the main road on the right.

There are still plenty of these terraces in the cores of the original settlements. In Llanfair they usually consist of between 3 and 6 small houses, presumably built by a single builder of fairly limited resources. The terraces are often not placed in any consistent fashion with respect to the road layout. For instance along the main Holyhead Road there are terraces fronting directly onto the road; at 90 degrees to the road accessed by a footway, and at various other angles. Given that there was plenty of land I speculate that the orientation of the terrace was probably most usually determined by size and shape of the parcel of land acquired by the builder.

In practice it is quite easy to pick out candidate terraces from aerial photos. I did a couple of passes over the buildings which had already been mapped on OSM and then cross-checked with old 6 inch maps at the National Libraries of Scotland and Wales (the Peoples Collection has more maps but no permalinks).

Modern day Llanfair PG shown on Bing Imagery. buildings identified as terraces are from OpenStreetMap and are shown in orange with a halo (to aid visibility).

Today Llanfair PG is a densely nucleated settlement. enclosed between the new and old Holyhead Roads, but certainly up until WWII, and perhaps as late as the 1960s it was a dispersed rural settlement. The location of housing terraces, which for the most part date from the late 19th century, highlights this dispersal. There are two main clusters at the two historical centres of the settlement, Pentre-Isaf and Pentre-Uchaf.

Memorial to villagers of Llanfair PG who lost their lives in the First World War. There is much of historical interest here, and personal one for me as my Grandmother's cousin is one of those listed, but it is the addresses which are germane to this post.

The War Memorial in the village is a useful source of historical addresses: virtually all are either individual house names or names of terraces. Amongst the terraces mentioned are:

• Crossing Terrace. next to the railway line and a named residential road today.
• Williams Terrace : off Ffordd Penymynnydd and behind Britannia Terrace.
• Bryn Goleu. Somewhere in Pentre-Uchaf, perhaps this building.
• Britannia Terrace. Fronting on to the Holyhead Road.
• Penucheldre Terrace. At right-angles to Holyhead Road and immediately to the W of Britannia Terrace.
• Min-y-ffrwd. Behind the Calvinist Methodist Chapel (and marked on the 6 inch map

Other places are easily recognised. Ty Capel & Ty Twr are both houses associated with, respectively the Chapel and the Tower (the Marquess of Anglesey's Column). Other locations are the local estates of large landowners (Plas Llanfair & Plas Newydd), and an island ([Ynys] Gorad Goch). My impression is that many men still worked on the large estates in a range of roles (farmhands, ostlers, servants, gamekeepers, etc., but 4 of the 26 had careers at sea. Others would have been fishermen (on Gorad Goch), dairymen or engaged in other agricultural trades. The silly long name on the station indicates that there was reasonable tourist trade which probably sustained shops and the family selling tickets for the Tower. Many younger people from the village would have moved to larger cities such as Liverpool or London to work in the building trades, in service or retail (all examples from my own family).

E end of Stryd Fawr (High Street) in Brynsiencyn. Two terraces can be seen in the middle distance on the right-hand side.
Almost all the houses on this street date from late-1800s.

I don't believe that there were any quarrymen in Llanfair PG. The adjacent village of Brynsiencyn did have a significant number of men who worked away during the week in slate quarries near Llanberis. Interestingly, Brynsiencyn is much more nucleated than Llanfair P.G. which may relate to having a development more greatly affected by industry. However, the main street there still consists of individual terraces.

Terraces elsewhere in North Wales

A similar pattern occurred in the larger quarrying towns such as Bethesda and Blaenau Ffestiniog. In Bethesda all along the A5, the main road through the middle of the town, are long terraces with up to 20 houses, such as Rhes Douglas (Douglas Terrace). Again each has an individual name and numbers belong to the terrace not the main road. In the suburb of Gerlan the quarrymen's cottages are clearly arranged in groups of terraces: some, such as Gwyernydd Terrace, being individually named on the 6 inch map. The same applies to the other great slate town Ffestiniog (some possibly developed by another great-grandfather of mine). Here many terraces are also named and it is clear that a fair number were only accessible by footpaths.

It wasn't only villages and quarry settlements where this held true. I first really appreciated this as a distinct phenomenon in 2010 when I carried out a brief survey around the suburb of Garth in Bangor, Gwynedd. Two roads connect the town to what was originally a fishing settlement. The upper road (Ffordd Garth Uchaf or Upper Garth Road) was largely developed in the 1930s with semi-detached houses, but the lower road, Ffordd Garth was built-up earlier. Largely this was as a series of terraces with 8 3-storey houses in each terrace. Again house numbering was within the terrace. Much more recently there has been an attempt to impose regular house numbers on both roads. House names have been supplemented by regular house numbers on Ffordd Garth Uchaf, and a single numbering scheme has also been adopted for Ffordd Garth. One can judge how successful it is by the fact that dustbins have the terrace name & number in the terrace painted on them.

Highways with "Terrace" in their name within Wales on OpenStreetMap
We can be confident that many have been missed as much of the detail mapping of Wales has used Ordnance Survey resources (either Out-of-copyright or Open Data)

At present OSM has at least 1500 highways (residential, footway, service etc) with Terrace in the name for Wales. An equivalent region of England, the East Midlands has less than 400, and the latter area has received more on-the-ground surveys than Wales.

Terraces Further Afield

A small terrace at Foxt, Staffordshire. Unusual in that one house was a pub.
Ian Calderwood on Geograph via Wikimedia Commons, CC-BY-SA.

Of course small discrete terraces are not restricted to Wales. A couple of examples come to mind: a terrace for brickyard workers on the edge of Maidenhead, and terraces in the Staffordshire village of Foxt close to the Churnet Valley which used to contain much water-powered industry. In addition in the densely developed urban areas of industrial cities another type of terrace was quite common. For the most part these have been demolished, but a few survive. In Nottingham these are usually two rows of terraced housing with a common walkway: typically they have 12 or 16 houses.

The Addressing Problem

Having provided some background on the history and distribution of this form of housing in England & Wales, it's time to look at why they present some issues for tagging addresses on OSM.

Loosely we can divide these terraces into 4 categories:

1. Those where the terrace name has been transferred to a standard residential road.
2. Those where some distinct vehicle access is provided, for instance by an unsigned service road.
3. Those accessed solely by a footway
4. Those fronting onto a street which has a different name from the terrace, and house numbering is only continuous in the terrace.

For the first three categories we can use the fairly standard Karlsruhe scheme with the name of the terrace held in addr:street. To ensure that Nominatim can retrieve such addresses in cases 2 & 3 I usually add the name of the terrace to the service road or footway (even when the service road only provides access to the rear of the properties. In some senses this is a fudge, but I think a justifiable one.

It is the 4th case which provides problems in the context of the Karlsruhe scheme. Essentially I know of two different strategies for resolving the conundrum:

• Store the name of the terrace in addr:housename and the name of the road in addr:street. (This is actually what Royal Mail do in their address file). 
• Add the name of the terrace as addr:street and make the houses members of an associatedStreet relation for the road.

Currently I prefer the second: at least it stores all the relevant relationships, even if they are not readily retrievable. My problem with the first approach is that it takes a common type of object and stores data associated with it in two different ways. We also cannot distinguish between a house which has both a name & a number and one which has a number and is part of a terrace. (Or a house in a named terrace with both an individual name and a number)

In general addr:flats is used for subsidiary numbers which share the same primary address. This distinction is needed because often a block of flats will both have a name and a number. It also has the advantage of not shoehorning different types of address objects into a single tag.

I'm sure that Frederik Ramm, the creator of the Karlsruhe Schema would say that the former approach should be used. The original intention was to store pure postal addresses after all. However, postal addresses are not the be-all-and-end-all of addressing. I've written before about how a Procustean approach to addresses, however convenient for the postal authority or company, is often unsatisfactory for a number of reasons. As the Karlsruhe Schema is what we have had for a long time on OSM, it is this which tends to be co-opted for other addressing needs.

The Mapping Problem

This post started out from the viewpoint of how to provide accurate addresses for this distinctive housing pattern. In the course of writing it I also realise that they also present a mapping challenge.

Land Registry Prices Paid ([not at all open less-and-less] openish) data suggests at least 2600 different street names containing "Terrace" in Wales compared with around 1500 on OSM. These might be missed because they are small groups of houses which aren't noticeable until a ground survey is undertaken; or they might be terraces aligned along main roads which actually require that the additional street name be collected (see above regarding terraces in Bethesda).

In Nottingham we know that many of the residual terraces were not named in OS Open Data products, and it was only around 2013 and 2014 that we identified a number in inner city areas (largely developed before WWI). Most of these were found in surveys conducted before our monthly pub meetings. Our usual meeting place, The Lincolnshire Poacher, was chosen in part because of transport links (it's close for buses to Derby, Mansfield and the Ashfields), but also because there is a very broad mix of different urban development within 10-15 minutes walk.  Many of the terraces within this area are not even obvious from residential streets, but come to light because there are odd gaps between streets or buildings which are not obviously accessible from the street, Field papers were invaluable for spotting many of these.

In Wales we don't have as that many mappers so they are less likely to be spotted as part of in-fill surveys. Also terraces are fairly common in small rural towns and villages as well as in the post-industrial parts of South Wales. We have additional resources in terms of open data and imagery now, so perhap sthis should be another mapping project for these times!