A recent discussion on the osm-gb IRC channel was about how to map chapels within crematorium buildings.

I thought it was worth summarising some of that discussion. These notes pertain to crematoria in England and Wales. I have attended a funeral at that in Geneva, but that was twenty years ago.

One of the difficulties is that most visits to a crematorium are likely to be to attend a funeral service. This is not conducive to any kind of sophisticated micromanaging, but does allow a decent amount of basic observation. In addition I’m not aware of well-developed tagging standards for various features, as may become apparent by looking at some of my examples

In general

In the UK crematoria started to be built after WW1 by local authorities. Burial space was scarce in larger towns and cities, especially in historical cemeteries and burial grounds. Initially, new large cemeteries were created by buying undeveloped land on the edge of the urban area. These included a central building containing the crematorium and one or more funerary chapels. Landscaping often created memorial gardens close to the building, and extensive tree planting (usually formal: avenues etc ) throughout the burial areas. The chapels are often recognisably religious type buildings (Landican).

Since about 1970 more standalone crematoria have been built, also by local authorities. By then older sales of burial plots were not providing adequate income for maintenance of cemeteries, and consequently a more sustainable financing was required. A number of private cemetery companies folded around this time including some of the “Magnificent Seven” in London. The response was to encourage more cremations and only lease burial plots for a number of years. Crematoria from this period tend to have more informal landscaping, more parking and more intricate traffic flow. The building itself is usually low and discrete rather than an obvious focal point.

Since 1990 quite a few more crematoria have been built, mainly by private firms (Denbighshire, Gedling). These tend to be in more rural districts, and may be more remote from urban development, and have less extensive grounds. The chronological listing compiled by the Crematorium Society shows this pattern clearly .

Other countries have quite different development patterns, for instance Carhaix-Plougher, which opened in 1988, was only the fourth crematorium in France. Now there are nearly 200.

Location & Directions

Crematoria are usually well-signposted from major roads. This is obviously necessary as visitors may be unfamiliar with the area. Some rural ones (Wrexham, Colwyn Bay) are located off quite minor roads. Others may be on a main road and do not allow access which crosses the carriageway (turn restrictions) as funeral corteges would disrupt traffic.

Older ones on the edge of town are likely to be directly accessible by public transport, but newer ones may require a walk from the nearest bus or tram stop, or just be difficult to access on foot (Gedling, Wrexham). Ensure that nearby public transport stops are marked, and map any pavements (sidewalks) along access roads. Where pavements are lacking consider adding verge tags.

The same roads should also have the most useful tags for good cycle routing added (any cycleways, cycle lanes, maxspeed, lanes, etc).

Access & Parking

The roads within the Crematorium site should be highway=service. Modern sites usually have a one-way system. These days there are roads which lead directly to the chapel entrance & exit (see further below), and these are solely for hearse and cortege (I’ve tagged them motor_vehicle=no, hearse=yes, but motor_vehicle=private works as well).

Foot access in newer crematoria is usually separate from motor traffic (Wrexham), but this is not true for older sites (Canley). I don’t know if any separate bicycles from motor traffic.

Parking is pretty much as one might expect, but take care to map accessible parking. Many funeral attendees will be aged and infirm. Some may have distinct parking areas for each Chapel, in which case name them. Older sites may just have parking along the main service roads (Landican, Canley), use parking=lane.

Footpaths from the parking area to provide access to Chapel entrances & exits should be mapped. Try to identify any steps (Bangor, Geneva) and provide info on step_count, surface, tactile_paving and handrails), and alternative step-free routes of possible. These latter are most likely on older sites.

Crematorium building

Most crematoria are built as a single composition containing chapel(s), offices, mortuary area and cremation facilities.

At Landican however the three chapels tower above the low connecting structures, with the cremation facilities behind the central chapel. This type of arrangement may be better represented by several buildings. At the very least distinct building:part elements can be used. For others the internal elements are best represented by use of simple indoor tagging (SIT).

There are also crematoria in much older cemeteries (Kensal Green, West Norwood, Cheltenham). At Cheltenham the crematorium is integrated with the original Victorian mortuary chapels; elsewhere these are on new sites withing the grounds.

Chapels*

Chapels can be mapped as place_of_worship with religion=multifaith. In most cases a node can suffice. Additionally, funeral_hall=yes can be used.**

Typically a chapel will have a waiting area, most modern ones with toilets. This may be an anteroom (Bramcote) or opposite the chapel entrance (Canley). Some anterooms have TV screens which may allow the room to be used as an overflow area if the funeral is well-attended (one can add capacity to each chapel). The chapel entrance is to the rear of the chapel. It should be marked on the anteroom if it exists).

Usually there will be a notice board outside the chapels listing which funerals are taking place that day (tourism=information).

Map toilet facilities elsewhere and any restrictions on access (e.g., Radar key holders at Landican). From this perspective it is quite useful to map nearby hospitality locations (cafes, bars, pubs) which could be used by people travelling from afar. (This similarly applies to places of worship used for weddings and funerals).

Usually a roofed area (building=roof) protects mourners arriving in the cortege and may be used by other mourners.

The chapel exit is usually to one side at the front of the chapel, and leads outdoors. Usually another canopy protects mourners as they convey condolences to the immediate family. Often there is an adjacent area for the display of wreaths and floral tributes (Wilford Hill). At Canley this is slightly removed from the main building and not protected from the weather (no available tag). The exit from xxx Chapel is away from the road and a covered arcade leads back to the cars.

Once entrances & exits are mapped the Chapel needs to be mapped as an area, either using S3DB or SIT. The choice depends on whether the chapels can be identified externally, and any need to map associated toilets and anterooms.

IIRC, the chapels at Geneva Crematorium are all accessed from a common hallway, and there is none of the detailed entrance/exit infrastructure.

Grounds

It is worth mapping various other features in the grounds around the buildings: notably other paths, hedges, but particularly benches. I have also mapped a substation (Canley).

A detailed survey of the grounds of even a small crematorium is a significant undetaking. There will be various memorial gardens, walls, hedges etc, with paths and benches. Do these first.

Trees

The biggest challenge will be the trees. In most locations the tree collection will be significant and contribute greatly to the overall ambiance. Crematoria and cemeteries often contain many unusual and interesting species and varieties of trees (see Steven Falk’s volumes on the Trees of Warwickshire for accounts of Canley & London Road). There are often lots of conifers, some of which may be hard to identify, and many Weeping forms of broad-leaved trees. If possible see if you can obtain access to the tree register for the location as this largely reduces mapping of trees to a verification exercise (say 30 secs per tree or tree_row, versus 2-3 minutes a tree without). Avoid mapping apparently closed canopy areas as natural=wood : this is nearly always wrong, often it just represents a few mature trees in a park-like environment.

Cemeteries in Nottingham have around 40 trees per hectare, which might provide a reasonable rule of thumb for estimating survey effort. Of these trees about a fifth are evergreen and one eighth conifers. I don’t have equivalent figures for a dedicated crematorium site.

Closing

This account is very much based on attending funerals over the past 20 years, often with elderly relatives with restricted mobility. It is very much a UK-based perspective, as Geneva is the only non-UK location I have visited as a mourner, and that 20 years ago. If you have a non-UK perspective, please take time to add remarks in the comments.

Equally, I’ve probably missed some things pertinent for UK sites. Please let me know.

Appendix : Sites Mentioned:

• Bangor*. A crematorium with a single chapel at the centre of a municipal cemetery built between the wars.
• Colwyn Bay (Bron y Nant). Crematorium and small cemetery tucked away under a hill. Despite being close the A55 Expressway access is not obvious.
• Wrexham (Pentre Bychan). A modern crematoriumwith a single chapel set in parkland dating from the 19th century, which is why I chose to use leisure=park. A beautiful setting.
• Kensal Green. One of London’s “Magnificent Seven” Victorian cemeteries. Created on open fields in the 1830s and therefore home to many plants which would not have survived either modern agriculture or urban development. The Western part of the cemetery, an extension, is still in use and a fairly modern crematorium is in use. Run by a private company, the General Cemetery Company. The original funerary chapels are located in the Eastern half of the cemetery, whilst the 1939 crematorium is in the extension.
• West Norwood. Another of the “Magnificent Seven”, but a municipal cemetery. As with Kensal Green, still in use for funerals and with a 20th century crematorium.
• Cheltenham. A crematorium was added adjacent to the original two funerary chapels. This closed recently and a new crematorium was opened in 2019 at the Eastern edge of the cemetery.
• Wilford Hill. A crematorium with two chapels at the centre of a 1930s cemetery built by Nottingham Council, but outside the borough boundary. The cemetery has a formal design.
• Bramcote. A modern (late 1970s) crematorium in its own landscaped grounds (developed from farmland). Current tagging raises the issue as to whether one tags the whole grounds (because cemetery tags are not appropriate) or the building.
• Canley. A mid-century cemetery with a crematorium at the centre. The cemetery is laid out mainly on a grid with many straight avenues, although close to the crematorium memorial gardens have a more informal layout. Numerous interesting trees.

• New private crematoria. Gedling and Denbighshire

• Carhaix-Plougher. A modern crematorium adjacent to the municipal cemetery.
• Geneva. The crematorium forms part of a large funerary complex and includes 4 chapels, and mortuary rooms. It is in the SE corner of the large St Georges cemetery. Currently mapped as several buildings when perhaps these are better as building:part.

  * This seems to be the word in use in both the UK and Switzerland, even though they are used for both non-christian and Humanist funerals. I’m a bit reluctant to call them funeral halls as this may conflate culturally distinct funerary rites.

** Note that whether to call these amenity=place_of_worship may well be culturally dependent.

In a comment on their recent diary entry publicerination suggested that use of leisure=pitch with sport=tennis was always intended for single tennis courts. I had my doubts given that someone in San Francisco was adding quantity, and I have added pitch:count and there is also a documented tag courts. Were we just the odd mappers out?

A relatively quick way to answer this was looking at tennis courts in the UK as I have a 1-2 year old import available for Great Britain. Overpass can also be used to collect the data, but it’s not possible to calculate areas directly. I ran a query which pulled tags, the geometry and the area of each pitch in square metres and saved this first as geojson and then as a csv file.

I then used R to build a simple histogram of counts of pitches in binned by increments of area of 25 sq. m. A tennis court is about 250 sq. m so this should show peaks at increments of this value. I also created a cumulative version of the histogram. Note these cutoff extreme outlying values most of which have been subsequently removed by edits by Dave F.

The simple answer to the query posed by the title is NO! Only about a quarter of tennis pitches represent a single court, with at least as many objects representing 2 courts and a similar number more than 2 courts. Whereas the first peak is tight, suggesting that single courts are mapped according to the visible court markings, the second peak is much more diffuse suggesting that these are mapped to the extent of the playable area (usually a wire fence). A tiny proportion (about 100 out of 20,000) have one of the tags indicating how many courts are represented by each area.

At least in the UK, tennis courts tend to be mapped together, and the data is not suitable for counting the number of courts. It is, however, suitable for finding places to play tennis which, from a conversation with the original author of Map at the weekend, was very much an original goal (albeit swimming pools in his case).

I’m all for using OSM for this type of analytical purpose, but it is important to look at how the data is structured rather than take a pure, and often personal, view of how it should be structured.

Coda: For some reason tennis courts has always posed issues as to how to map them: single courts or in groups; the in-play area within the tramlines, or the entire playing surface, etc.

For a number of years I have sporadically looked to resolve missing speed limits on major highways (highway=trunk and highway=primary) in the UK. I use a simple Overpass query to find these in a manageable area (usually former regions of England + Wales, Scotland and Northern Ireland.

The other day a discussion arose on IRC about some distinctly English placenames in North Wales. Looking at the area I realised that the A493 had no speed limits added. As both Mapillary and now Bing Streetside imagery are now available in the iD editor, I was able to add limits from Aberdyfi to Tywyn. In the past this nearly always required a survey.

I was going to write about looking at limits elsewhere, but found it easier to add various pictures in a short Mastodon thread.

tl;dr: There’s lots of work everywhere to improve the mapping of maxspeed on highway=primary.

I started drafting this after b-unicycling’s report on her trip to Anglesey as it reminded me that when mapping solar power on the island, I’d noticed a lot of old windmills.

In most of Wales there were abundant sources of water power. So water mills were common before steam engines were available. Many were corn mills, but woollen mills were also common. There was even a tidal mill at Carew.

The only one of these Anglesey windmills I knew about beforehand was the one on Parys Mountain. The stump of the tower is visible from afar. It was used for pumping water out of the copper mines (at one point the largest in the world).

Given that I’d noticed several, most of which could be found marked explicitly as windmills on old maps, I wondered if I could search for others more systematically. Fortunately, the Cymru 1900 project had transcribed text from old maps around the start of the 20th century. In fact the original project, an initiative of the National Library of Wales and other heritage organisations, was expanded to cover the whole of Great Britain. Searches of specific terms can be made from the NLS website, or the whole database can be used..

I downloaded the windmills data (as csv) and loaded it into Josm. Selecting just those in Anglesey gave a simple set of targets to check with the To do plugin.

A few are still working, but most have been converted to residential accommodation, or the tower has been secured (as at Llangefni).

I’ve been a bit naughty in tagging these with disused=yes rather than using a life cycle prefix. The main reason for doing so is to ensure visual feedback whilst editing. (It would be very nice if one of the OSM renderers offered a disused windmill icon ( as the Ordnance Survey did in their 7th series map).

Anglesey has always been grain-growing country, so the presence of corn mill’s makes sense. A maritime climate, allowing growth in the winter months, and a relatively benign topography, particularly compared with the rest of North-West Wales, undoubtedly helped. To the mapper, scrutinising buildings on aerial imagery, cattle farming is actually more apparent these days.

Undoubtedly the text extracted from the old maps from 1900 can be of use in searching for other features, and not just those of historic interest.

The OSM-UK has previously discussed using “F.P.” as a means for locating potential and actual public rights of way. Christian places of worship (those of other faiths were thin on the ground in 1900, although the first mosque in Britain already existed), for instance may help filling the mapping gap for places of worship (the quarterly project for Q4 this year).

I don’t know if they’ll be any help for b-unicycling’s quests for crannogs, tumuli or duns. That is for her to find out!

In the UK and Ireland stone circles are amongst the oldest built structures. They are usually mapped with historic=archaeological site, archaelogical_site=megalith. megalith_type=stone_circle. However, mapping in Wales it is soon apparent that there are quite a few modern ones.

Most are Gorsedd Stones, relics of National Eisteddfodau. However there’s a small covey near the entrance of the show cave at Dan yr Ogof. Most are mapped as monuments or memorials, although I think when the Eisteddfod returns to a site they are used for their original purpose.

Modern stone circles exist elsewhere, sometimes associated with the Welsh diaspora, as at the Plaza de Colonnes in Gaiman, Argentina (see image at top), and possibly in Hungary. Others, without a Welsh connection, include: replicas of Stonehenge (in a variety of materials) in the USA, Australia & New Zealand; and one in the grounds of the Glastonbury festival site.

Modern stone circles represent a type of tagging issue - the most widely used tag does not truly encompass all instances - which is not uncommon. Often these can be resolved by adopting a second more generic tag: one which is implicit from the first tag. A good example are disused telephone kiosks reused for other purposes. amenity=telephone describes a public pay payphone, but implicitly suggests a kiosk or booth (sometimes tagged as such using booth). Disused ones housing defibrillators, book exchanges etc. are often now tagged man_made=telephone_box **. Restaurants and bars in hotels are another example: often tagged bar=yes or restaurant=yes, particularly if these facilities are solely or predominantly for hotel guests. Minh Nguyen maintains a range of wikipages dealing with tagging issues of this kind related to synonyms, homonymy and antonyms. Although these are often easy to understand as a person because of context, they often create complications for interpretation by software, especially editors.

In this case I suggest man_made=stone_circle as a suitable encompassing tag for both ancient & modern ones. I’d also like to be able to map the stones. In some circles, henges and alignments the individual stones are themselves megaliths (also a historic tag), but there are many circles, whatever their age, where the stones are quite modest in size. ‘Stones’ may also be a misnomer as some Stonehenge replicas are made from a variety of other materials.

These mapping wants are a quirky little thing, but worth exploring. As here, modern versions of historical objects, whether replicas or re-imaginings (such as the development of commercial mausoleum in the form of long barrows) are probably the single largest category. However, there are others:

• Military vs sport shooting rifle ranges. Many of the latter may be mapped as leisure=pitch with sport=shooting, but this is not only misleading, but fails to provide for both indoor ranges and many outdoor ones such as those in Saas and Kublis. An added complication is that some sport shooting ranges are fairly closely integrated with military ones (ranges are so common in Switzerland because male citizens need to maintain & demonstrate continued proficiency).
• Estate agents, usually mapped with an office key, but there is a big difference between estate agents located on the high street in a retail property and one in a commercial district.
• Milestones has been accepted as the generic tag for all distance markers. However, there are both historic & modern examples. Obviously the original intent was to mark the former.
• Other historical examples : wayside crosses, tumuli and other burial mounds.
• Various types of marine rescue operations and facilities. Currently subject of a proposal vote.

There are three approaches:

• Use different keys, but often with the same tag value
• Subsume all objects under a single key-tag pair, perhaps using sub- or supplementary tags to distinguish types of objects. Thus modern instances of a typically historical object would still receive the historical key, but will be marked as different through another key.
• Treat the common base tag as a special case of another tag value which is implicit. This, in particular, is true of many things tagged with the historic key, where one might consider there is an implicit tag with the man_made key: in this instance man_made=stone_circle.

All of these will work for regular data consumers, but in the second case agreed subtags are important.

Ad hoc data consumers are less likely to be aware of their existence, or how they can change the base tag meaning, sometimes subtly, but in other cases drastically (as sport=shooting does on pitch). Probably, the choice of approach will depend on the nature of the tags, and the likely dissonance felt by map and app users: not sure that I’d be very chuffed it I went out of my way to visit a historic stone circle to find it was some blocks of concrete put in place a few years ago. Similarly if the remains of a loved one are interred at Soulton Long Barrow, I’m not sure I’d be pleased to see it marked on a map as an archaeological site.

** SomeoneElse points out that there are [many other ways)](https://github.com/SomeoneElseOSM/SomeoneElse-style/blob/master/style.lua#L1782 of indicating a telephone box or kiosk.

Lord’s Rake: one of the steepest paths marked on OSM in Upper Wasdale

In my last diary I introduced the idea of using elevation models (DEMs), specifically a DTM (terrain model) to find sections of hiking paths on OpenStreetMap which may cause problems for regular hikers. In this sequel I describe a refined approach using a higher quality terrain model and a vertex-based approach to calculating slope angles actually likely to be experienced by walkers.

Wasdale steep paths: larger image at 1:25k

Fortunately in most the UK we now have high quality recent Lidar data covering much of the country at a resolution no worse than 2 metres. It is a bit tedious to download as it is available in tiles 5 km on each side (names are based on the 10 km grid square plus a compass direction, e.g., SK53NE), so instead of covering a 100 km square as in the previous example I have just focused on one 10 km square NY20. This covers many of the most difficult paths in the Lake District: those around Scafell, Scafell Pike, Great Gable, the Langdale Pikes, and a bit of the Helvellyn range. Unfortunately not all the Lidar data is 1 m, but the area of Upper Wasdale is thoroughly covered.

Basic Workflow

Lidar tiles are downloaded and unzipped into a single directory. I use QGIS to run a gdal command to create a virtual raster which combines all these tiles into a single file I can load into QGIS. This approach is particularly useful as the virtual raster parameters allow merging data at different resolutions.

From the DTM I then created a hillshade layer and a slope layer. Both of these were more to help with visualisation, than to do the work, which used the base DTM.

As in the previous example I used the QGIS processing function “Reclassify by table” to create a second version of the slope layer with slopes of greater than 25 degrees grouped into 5 degree buckets. This layer is saved as a file for future use.

OSM path data is then loaded into QGIS, and all vertices extracted. Each vertex inherits all the fields of its parent way plus a vertex index. In practice I had to do a bit of data wrangling because path junctions would create multiple vertices in the same place.

I downloaded and installed the “Point Sampling Plugin” for QGIS and ran this to obtain height values for each path vertex. At this point I had to do some wrangling with the point data to merge these values back with the original data, but I suspect I missed a feature.

At this point one has each point in the path with where it is along the path and how high it is above sea level. I took this data and pushed it into PostGIS.

With PostGIS one can use a PostgreSQL window function to rebuild the path segments and to calculate the height difference between successive nodes and the distance between them, and consequently the slope of the path.

The resulting data is then pulled back into QGIS where it can be filtered and plotted.

Results

There were some immediately obvious improvements. Firstly, this approach eliminated the obvious false positives of shallow angle paths traversing steep ground. Secondly, there were many fewer high angle outliers. The vast majority of steep path segments are between 25 and 30 degrees, with only a few as much as 35 to 40 degrees (including part of Lord’s Rake, shown at head of post). Thirdly, most of the paths picked out very much match my expectations.

There are small number of extremely steep sections (for instance exiting the West Wall Traverse), but I suspect many of these are an artefact in that an OSM path being only 2-3 metres out of the correct alignment may move it onto a cliff face. An unanticipated aspect of this is the potential to improve OSM mapping in terrain which is difficult to see on aerial images.

The particular paths I expected to be picked out were. A few examples: the Climbers Path under the Napes on Great Gable, most paths around Mickledore, Brown Band on Great End, the descent from Yewbarrow to Dore Head.

A few paths stand out as worth checking:

• Above right bank of Piers Gill. This is noted by the local mountain rescue as being an “accident black spot”.
• Short cut avoiding Broad Crag Col. This is shown on OSM crossing a couple of cliffs, and is probably misaligned anyway. The data shows it have a couple of very steep sections. This looks to be a route over scree, which may be suitable for descent for very experienced walkers, but is unlikely to be of any appeal on ascent. (One note, in snow many more routes can present themselves, as I found way back in April 1984).
• Direct route to Yewbarrow. I can’t remember seeing a path in this location and have always taken the route which more-or-less follows the south ridge from its base. It is visible on imagery, so it must have arisen relatively recently.

One path (used advisedly) which is not marked on OpenStreetMap is the route descending directly from Dore Head to Mosedale. This roughly aligns with this mapped bit of scree. Parts of the scree have completely washed away and it is no longer a straightforward scree run. Again it is a common location for accidents. Certainly, no longer a quick way to get to the pub.

Conclusions

This has been an interesting exercise for me, as I’d never really done a combined raster and vector analysis in QGIS. There are still a few steps which are a bit cumbersome. I suspect most could be overcome by using PostGIS for associating pixel values from the raster data with vertices of vector data.

There are obvious advantages in being able to analytically create a set of data for checking rather than relying on non-systematic sense checking by individual mappers. It’s already been suggested this is also useful for things like advocacy with the bodies who manage trails elsewhere. The data could also be potentially used to enhance extracted OSM data for hikers.

My next steps are to : a) complete DTM coverage of the NY20 hectad; b) progressively enlarge the vrt file to cover the whole Lake District; and c) to repeat the analysis for the entire area.

Recently there has been quite prominent press coverage of mountain rescue incidents in the English Lake District involving people using various outdoor activity apps (The Guardian, Grough). It turns out that these incidents involved paths mapped on OpenStreetMap, and have been discussed by the local UK community.

In general the Lake District is a fairly benign environment for walkers with the bulk of the upland paths being heavily used and for the most part not difficult (SAC Scale mountain_hiking). A number of popular walking routes are well known to be more difficult : Striding Edge, Sharp Edge, Lord’s Rake, Jack’s Rake. However, there are other much less well-known harder trails and scrambles. As OSM data gets richer, the more likely such things will get added. In addition, contributors may be tempted to add routes that they have used which either have traces on the ground which are either vestigial or none existent. It’s also possible that paths have been added from old maps, or even from poor quality GPS traces.

I therefore thought it might be useful to try and scan any walking paths marked on OSM within the Lake District which may be problematic. The approach I have taken is to find paths with a steepness of 25 degrees or more.

Here I describe my first pass attempt, which has various deficiencies, because the elevation model I have used is quite coarse (25 m elevation pixels).

To determine steepness I downloaded the basic terrain model from Ordnance Survey Open Data : Terrain50. This comes in a range of formats and packages including raster DEM and contours. Some of these are zipped files containing multiple directories themselves containing zipped files. Most data is stored at the basic level of a 10 kilometre grid square, known as a hectad (SK53 is one such).

I downloaded all DEMs for the NY major grid square (which includes the Northern two-thirds of the Lake District. I then built a virtual raster encompassing all one hundred tiles in QGIS and saved this as a .vrt file. From this one can use basic raster tools in QGIS to do various manipulations of the data: hillshade, aspect, slope, contour extraction etc.

For this I created a slope model. I then used one of the “Processing” tools “Reclassify by Table” to assign slope values over 25 degrees to buckets split by 10 degree boundaries. I then converted the reclassified raster slope data into polygons.

I extracted all ways with a highway tag containing path, track, footway, bridleway or cycleway within the Lake District National Park using Overpass. Within QGIS I then clipped these by the polygon layer representing steep ground, to obtain a basic set of paths to examine. Note this does not examine the steepness of the path, so a path which is pretty level and just traversing across steep terrain (e.g. Wasdale Screes) still shows up. Also the coarseness of the DEM inevitably catches paths on top of, or at the foot of, cliffs. I will therefore do a second pass with : a) a much finer DEM (Environment Agency Lidar, 1m); and b) calculate actual path steepness with heavily noded paths.

In order to split the paths by the steepness of the terrain I loaded both the steepness polygons and the clipped paths into PostGIS. I had to repair a number of the polygons, but this allowed me to split the paths into sections each labelled by the apparent steepness (using the lower bound of the range, so a angle=30 means between 30 and 40 degrees). The image at the top shows paths around Scafell and Mickledore, including Lord’s Rake.

I’ve put the geojson of these path sections on umap if anyone wants to scroll around. Undoubtedly, tagging can be improved on some of these.

PS. This is very much a “quick-and-dirty” first attempt. The proper way to do it looks to be to assign elevation values to individual vertices of the paths, but that then requires reassembly of the path for visualisation purposes.

A few days ago i provided an example Overpass query to show buildings with a mapped start_date colour coded by age. This was in response to a query by long-time Latvian contributor richlv. Another user based in Latvia asked on Mastodon if it was also possible to look at data by how long ago since it was edited.

This proved to be quite a lot harder than my previous example. The issue is that the “@timestamp” field in Overpass-Turbo is always treated as a string and is never cast to a number or date. This meant that the MapCSS queries have to deal with regular expressions, so I’ve just done the bands in years (“way[@timestamp=~/YYYY.*/]”), as I haven’t experimented with how rich the regexp implementation is for MapCSS. An example of the amended query for roads and buildings in a given bounding box is here.

If this seems vaguely familiar, ITO used to provide a tool, OSM Mapper, to visualise data by age (and many other parameters). Neither this, nor their later, more broadly-based, mapping tools ITO Map have really been replaced.

One, other thing, Keir Clarke at (Google) Maps Mania showed how to achieve the colour coding I mention in MapBox Studio (using data from Lviv). This slightly misses the point of doing the work in Overpass Turbo: it would have been much easier for me to do it, for instance, using QGIS, but that is no help to someone who wants a quick answer without using a toolchain (QGIS, MapBox Studio, TileMill, Kosmtik etc). The OSM ecosystem is still quite weak for users who want a relatively simple solution within a web browser.

The other day richlv asked on IRC if there was any OSM-based rendering showing the age of buildings. Although I could think of a couple of examples where people have done this, they did not use OSM data (other than the extremely early work.

I made use of open data of buildings from Portland Oregon to look at clustering, but the inspiration, and awareness that the data existed, came from a MapBox blog post.

A similar approach was taken by Waag who made use of the BAG open data on buildings for the whole of the The Netherlands.

I wondered if it was possible to use MapCSS styling within OverpassTurbo to create a simple way to achieve the same effect. After a bit of experimentation I was able to do this. I used Dutch localities as test areas as all buildings have been imported from BAG and therefore have start_date tags in a consistent format (“yyyy”). I also looked at other places with some buildings (usually those with a heritage protection) have start_date tags.

A couple of things I found:

• Tagged nodes (e.g., entrances) on a building outline got the default render. It was therefore necessary to have a MapCSS rule which makes all nodes have 100% transparency.
• Rules of the form [year<1925][year>=1950] didn’t seem to work as expected. Instead I relied on a single predicate and ordering the rendering rules so that the rule finding the oldest building is the topmost.
• I couldn’t find a way to both use the date() parsing function in Overpass and retain the original geometry. This is a shame because it should handle cases such as “c1860”, “1861-1862” and “1861-07-14” which a simple-minded approach does not.

All of these may just because I’m not familiar enough with these aspects.

Once I had some basic principles worked out, a big bonus was that the Waag Carto-CSS style sheet is available on Github, so I was able to copy both the precise colours they used and the date ranges. The resulting Overpass-Turbo query used to generate the map at the head of the post is here. Note that I used the Carto-DB dark tiles to provide a sufficiently dark background, and you need to be sufficiently zoomed in to see all buildings (if small POIs are not shown as points).

Just to show it working elsewhere, here is Central Manchester, with pretty much the same buildings mentioned in Frankie Roberto’s talk of 2009:

Back in March I was amused to see Amanda tweet that the UK OpenStreetMap community ran a solar power mapping project “several years ago”. This was the theme of the [quarterly project(https://wiki.openstreetmap.org/wiki/UK_2019_Q3_Project:_Solar_Power) in Q3 2019, but the project keeps trucking on.

All small-scale Solar Power in Wales (i.e., excluding solar farms of more than 1MW capacity)

In the past few days we reached a milestone of having comprehensive solar cover for Wales, one of the constituent countries of the United Kingdom. I think this is the first country to have solar power mapped at such a fine level of detail either on OSM or anywhere else.

The rest of this post discusses aspects of the how & why of this work.

Solar Mapping in the UK

Active mapping of solar farms had been proceeding for a while prior to 2019, but it was Jack Kelly of OpenClimateFix who posed the question about mapping rooftop solar photovoltaics. His problem is nowcasting changes to power generation from solar which could reduce the need for some carbon-based standby power. When clouds pass over, solar generation can drop, leading to transient changes at a local level. Rooftop solar is a not insignificant part of the mix & both highly dispersed and clustered (at all levels: groups of houses, streets, estates, local authorities and even countries).

Current Status

Solar farms were pretty much all mapped by the time Dan Stowell talked about the project at SotM-19. Ongoing mapping is mainly about rooftop solar, with a much smaller number of ground installations, ranging from 4kW to 300kW.

We reached 40% coverage for Great Britain at the end of 2021, 400,000 individual solar panels in January this year, and currently there are 449k panels, 396k installations (an installation can have more than one panel) representing 46% of the total government figures (or 39.6% if we use a higher estimate of 1 million installations). Mapping in Northern Ireland has been more limited, but VictorIE has made decent inroads recently.

Over the 3 years since we seriously started mapping rooftop solar we’ve benefited significantly, in terms of visual feedback from Gregory Williams website and Russ Garrett’s OpenInfraMap. It has also been a good activity during the pandemic when regular mapping has been curtailed. For some of us this is still the case. Aerial imagery has also improved significantly, with installations going form black blobs to ones with clearly delineated individual modules. We can also use property boundary open data to align imagery more precisely.

At the outset we tried many different areas, but ended up focusing on a relatively small number with good aerial imagery & readily observed solar. Even then it was only possible to exceed 50% coverage in a few areas. Particularly noticeable was that visual search in rural areas was pretty hit-and-miss. I personally gave up on Anglesey, having only found 20% of the installed base and most of that in one estate in Llangefni.

Eventually, after a bit of trial and error with various open data sources, I found that clustering buildings from Ordnance Survey Local gave a small enough set of areas to search. A geojson of centroids can be used either in the Josm “To Do” plugin, or the Potlatch 3 task option. Using a Lower Super Output Area (LSOA) as a unit of work proved convenient as most can be searched in under 30 minutes, and as this is the lowest reporting unit of Gregory’s map, feedback is available in 24 hours. I first applied this in parts of Devon, but had long wanted to try it in Powys (the most rural area of England & Wales).

Overall Goals

Our initial goals in mapping solar were therefore:

• Investigate the practicability of mapping with OSM. Initial results were encouraging as the current status confirms.
• Map enough solar to allow investigation of nowcasting and other predictive approaches. This required large numbers of installations to be mapped, unlike many other OSM use-cases where usefulness either does not have a threshold, or the threshold is much lower.
• Document preferred building types used for rooftop solar. This is possible, but probably requires a better attributed & more consistent set of building data than currently available in OSM.
• Enable the creation of a solar cadastre, as was carried out with the OpenSolarMap project of Etalab in France.
• Document a significant change in the appearance of both urban & rural landscapes.
• Provide enough data in specific areas for research & case studies around domestic solar power. A comprehensive data set for Wales represents a big milestone in this regard.
• Provide good training data for automated recognition of roof-top solar from aerial imagery.
• Automated recognition of solar from aerial imagery. Although Tyler Busby did some initial work on this back in 2019, this is probably the one goal where I feel we have not made as much progress as I had hoped.

The current Project : mapping solar in Wales

Last December proved to be a convenient time to return to mapping in Powys. I spent that month adding solar, and a few other things, to Powys. This proved to be productive and met my initial goal both with respect to comprehensiveness (see below regarding comprehensiveness) and level of detail (q.v.)

Having got around 85% completion of Powys, my thoughts turned to the other rural areas of North Wales, Gwynedd and Anglesey. I started on these two in the New Year, and before I knew it realised I’d embarked on covering the whole of Wales. Originally, I anticipated this would be a six-month project. However, a few things derailed that timescale:

• City Nature Challenge 2022: without Muki Hakaly’s prompting I might have missed this after participating for the previous two years. This in turn led me to spend a lot more time recording nature on iNaturalist & other platforms.
• Degraded imagery. Bing imagery started being replaced whilst I was mapping Cardiff in April. The new imagery is much less well orthorectified than the old and is considerable more oblique. This made the process of mapping & tagging panels rather more tedious, and meant that many more installations needed close attention.
• Well-mapped areas needed more time. The councils which were already at 60% (basically along the N. Wales coast and Wrexham) took rather longer to work though than I anticipated. Similarly earlier work by brianboru, Gregory Williams, ZenPhil, and myself often required not just adding module counts & direction, but repositioning.

However, I did reach my target about a week ago, just dipping into August, making 8 months altogether.

Apart from being a country in it’s own right, Wales also encompasses a very broad range of urban and rural areas from dense cities and post-industrial towns to sparsely populated valleys with a few scattered farms in upland areas. It’s therefore a decent microcosm for mapping solar elsewhere in Europe.

Next Steps

Of course there is still a lot to do with the Welsh data. Some basic tidying up will be the necessary over the next couple of months;

• Convert any remaining larger solar mapped as nodes to areas.
• Ensure all larger rooftop solar has a building underneath it.
• Review some LSOAs where less than 70% of solar has been captured. Often imagery has been updated.
• Split or add buildings where the algorithm used by Gregory to calculate installations either under or over counts installations. (I have had to do a lot of work with the Terracer plugin in Wrexham for this reason).

Other people are mapping buildings, for instance around Merthyr & Rhondda Cynon Taf, so gradually this mapping will also improve solar by improving the level of detail, but ultimately remaining buildings will need to be added, so this may be the next major project.

Further afield, the use of clustering of buildings can be used anywhere that OSM has a decent number of buildings: the OSM-IE buildings project & OSM-FR mapping from cadastre both offer opportunities to improve solar mapping now.

Appendix: Some Terminology

Comprehensive vs. complete

I use the term comprehensive rather than complete. We can rarely claim any subset of OSM data is complete, but it is not unusual for the data to reach a level where it is quasi-complete (i.e., perhaps missing a small number of features, and requiring some updating). Limitations to achieving true completeness are : continual change; use of aerial imagery which is a few months to a few years old; difficulty of interpretation of imagery; ordinary omissions & commissions which occur in mapping (fortunately much less than 1% in this case). Good examples are the highway network, where typically we miss new residential developments and may not update changes in one-way systems., or mapping shops where change is continual.

Wrexham : a comparison of how relative complete mapping (80%) compares with comprehensive mapping (at 96%). Note greater dispersion of elements showing a more successful search approach.

My initial target for comprehensive mapping of solar was to achieve at least 80% of the known installed capacity, but with one or two exceptions well over 90% has been reached. Continued mapping & roll-out of new imagery, and even a bit of on-the-ground surveying can then incrementally increase this (as can be seen in places like Watford & Canterbury).

Level of Detail for rooftop solar

In terms of Level of Detail (LOD), there are perhaps 5 stages of mapping rooftop solar:

1. A node tagged with location, locational accuracy perhaps 10 m root-mean-square (rms). (This proved the most practical LOD for all our initial mapping. Vagaries of aerial imagery quality, alignment & recency made it too hard to do extra detail, plus the optimum zoom for finding solar does not resolve detail
2. A node with additional tags for direction & module count ( a proxy for power rating), but also more accurately located (5 m rms). An area without these detailed tags is equivalent
3. A level 2 node with an underlying building, ideally topologically accurate, but certainly slightly more accurate location (rms 2.5 - 3 m). A specific issue is that in some places building locational accuracy itself is in the order of 3-4 m rms ( buildings had been mapped before the cadastral data was available to improve alignment)
4. As for level 3, but converted to an area, building alignment refined.
5. As for level 4, but the underlying building is mapped with S3DB.

My goal for Wales in the current project was to have everything mapped to stage 2, and all ground and larger rooftop installations (more than 30-40 modules) at stage 4. This will still leave many thousands of buildings to map to further improve the level of detail for ongoing work.

A few days ago a few mappers in the UK noted that Bing imagery seemed somewhat out-of-date. I noticed it because it appeared not to show recent housing developments until zoomed in at z20. I found a development just outside the village of Llanbedr Dyffryn Clwyd whilst mapping rooftop solar in Denbighshire. My workflow for solar mapping uses Josm. As all the new houses had solar panels I wanted to add the buildings themselves, and I find the way in which imagery can be aligned easier to use in iD than in Josm (particularly as the offset needs cancelling in Josm which does not fit task-based mapping over several thousand square kilometres). So I did it a little crudely in Josm, only to discover that at z20 the imagery was available in iD, so I tidied things up a bit.

I didn’t think much about it until a local mapper in the area commented that Bing seemed a bit behind. This seemed a bit more significant, so I looked at my local university campus which is under continual development and therefore has lots of features which enable one to age imagery. To my consternation the zoom levels down to z19 showed a building which was demolished at least 7 years ago. Whilst looking at this area just now, the z20 imagery appears to be disappearing from cache.

A short distance South of the University is the Nottingham tram line which was well under construction in 2013 because a major bridge was put in place in September of that year. The ‘updated’ Bing imagery now pre-dates the tram development, and any construction work on the Chinese Studies building which opened in January 2013. It does show a new lecture theatre on the main campus which was built post-2009. It therefore appears that the imagery has reverted to a state around 2011 or 2012. Co-incidentally or not, this seems to be the same as ESRI Clarity.

I’ve looked at a couple of other places where I know construction work bridges this period:

• CMU Campus N of Forbes Avenue, Pittsburgh. A lot of new buildings replaced a car park and Forbes itself has been extensively re-modelled to incorporate bike lanes. Bing imagery at all zoom levels seems relatively recent.
• Hardbrücke, Zurich. The former terminus balloon loop of the Number 8 tram has changed and the tram now enters and crosses the bridge (opened in 2017 IIRC). Bing imagery pre-dates the tramline extension.

Several others have noticed similar changes in the UK, and there is a relevant Github issue for the iD editor. The deterioration in usefulness of the Bing imagery appears to be associated with a recent Pull Request designed to eliminate problems with OSM editors exceeding a daily allowance from Bing. Regular Bing Maps shows much newer data, e.g., at Hardbrücke and the QMC tram stop.

I’ve posted a few screengrabs of these location on Imgur

I don’t know how long this will take to resolve at the code/organisation level, but I think it is important that people editing be aware that Bing imagery may be considerably older than they may believe. Clearly at this stage it has not inconvenienced too many people editing. Bing Imagery in other editors is not affected.

PS. The Bing imagery in other editors is largely recent Maxar imagery and was updated around the end of April. In many places across Wales this new imagery is poorer than that which preceded it (true also in Nottingham, where I could see the works for a new underground 11kV cable which I watched being installed).

For the past few weeks I’ve been making a concerted effort to map solar panels across rural areas of North & Central Wales (Powys, Gwynedd and Ynys Mon - Anglesey), so far with good results. I’m using a thorough search technique which looks at individual clusters of buildings from Ordnance Survey Open Data (which is complete for Great Britain). This means I see lots of other things which need mapping, but from experience I know it’s important to focus on the specific task. However, I have followed up some of the more striking things, which I plan to report in a series of posts.

First up was a striking structure in the middle of farmland on the Llyn Peninsula. It was pretty obviously a high-tech milking machine: milking parlour seems a bit quaint for a pretty sophisticated bit of kit.

Bing Imagery (close-up view).

Bing Imagery (context).

Fortunately there are some good Geograph images which confirmed my intrepretation, and provided other information too.

The Cefnamwlch Home Farm Milking Parlour

One of these says the dairy has over 1000 cows, so the machinery is probably working pretty much all the time. When seen in context it is apparent that there are wide trackways converging on the parlour allowing continuous streams of cattle to flow and from the site.

Entry to the Milking Parlour

Super-sized dairy farms are still relatively uncommon in the UK, and most probably keep the cattle under cover all the time (as in the US), so this one with the milking machinery open to the air is even more unusual. Subsequently, I’ve found one other on Anglesey at Bryncelli Ddu outside Llanddaniel Fab.

In both the footprint of the farmyard area is substantial, but most distinctive is the engineered network of trackways (sometimes including underpasses or overpasses to avoid minor roads) which also occur on farms with indoor milking parlours, such as Plas Llanfihangel (also on Anglesey). Given the continuous to and fro of cattle, and the associated muck, I presume these are solely dedicated for their passage.

I’m not sure that there are any useful mapping points to be learnt here, as most milking parlours are just non-descript large farm sheds, and tracks are tracks. However, there are other distinctive farm buildings which are associated with other intensive husbandry practices which I’ll look at soon.

Whilst writing my previous post on Welsh Heaths I discovered that a hill on the Llŷn Peninsula shown as heath in the Phase 1 survey has no landcover mapping on OSM. I walked up Myndydd Rhiw in 2008, just before I started contributing to OSM – so I failed to take enough photos. The weather was not very good that day so I’d opted for a car journey to visit locations on Llŷn rather than brave the rain in Snowdonia. Rhiw was the only place which wasn’t chosen because of personal associations: not only had we holidayed there several times during my childhood, but my great-grandmother’s family come from Llŷn,

Summit of Mynydd Rhiw

The Llŷn Peninsula consists of old rocks, those on the S side being Edicarian in age, with bands of Cambrian Rick’s along the North Coast, the two being separated by a fault. Intruded into these during the Caledonian Orogeny in the Ordivicaian are isolated groups of hills, mainly of granite. Mynydd Rhiw is the most westerly intrusion and is mafic (more basic than granite). These outcrops are distinctive not just from their height above the gently undulating farmland of the rest of Llyn, but also because even at a distance they obviously support heath vegetation, as seen in the photo below.

View NE from the road under Mynydd Rhiw

I’ve used a few suitable photos from that visit to show an example of “Dry Heath” from the Phase 1 classification I described in my earlier post.

Some, but not all of the open area on top of the hill is owned by the National Trust, as can be seen on their land-holding map.

Although I have my own recollections and photos and the Phase 1 habitat data, actually determining the boundary of the heath polygon is not straightforward:

• The Phase 1 data is clearly out-of-date. The patch of forestry to the E has changed shape, probably through clear-felling of substantial parts of the area.
• Phase 1 data indicates Western Gorse Ulex gallii as a common dominant or co-dominant with Bell Heather Erica cinerea. Unfortunately, I did not really look at the vegetation on my visit, and Western Gorse is very low growing, and flowers at a different time to Common Gorse so is not obvious in imagery. However, there is one excellent photo showing it in flower on Geograph (unfortunately not precisely geolocated), and a couple of others from the same source showing the ecotone between pasture and heath on the W side of the hill.

• On Bing Imagery and in my photos most of the top of the hill consists of unbroken areas of heather with, presumably, mown grassy strips in between. This pattern is not consistent between the various imagery layers (see examples below). In the absence of any knowledge about how & why it is managed the actual status is guesswork. My suspicion is that the heather is cut back periodically, possibly to avoid encroachment by bracken.
• Two heathland codes are used in the Phase 1 data: D1.1 for unbroken heathland and D1.5 for a mosaic of heath and grassland. This latter code may be appropriate for the pattern described above. I just don’t know.

On the other hand the entire area to be considered is quite clearly delineated on aerial imagery. The status of former forested areas cannot be discerned without a visit, but these are peripheral to the site anyway, so can be left for someone else.

Bing Imagery

Mapbox Imagery (presumably from Maxar as that layer & ESRI look similar)

The changes in appearance of the heath area meant that I no longer felt able to use the Phase 1 polygons directly.Instead I used them as a guide for splitting the heath area into several polygons, largely aligned with the stone walls and property parcel boundaries. The major divisions were between the top of the hill which in all imagery shows heath vegetation, and the sides of the hill which seem to show variable amounts of it (probably with a good admixture of stands of bracken).

The purpose of looking at Mynydd Rhiw was to use better evidence to map a small well identified heath. Instead I show that even with goodish information high quality mapping really requires a recent on-the-ground visit.

If you do fancy visiting, the views south to Porth Neigwl (“Hell’s Mouth”) are first rate!

A couple of weeks ago a research group in Oxford published a worldwide dataset of predicted solar power locations in the journal Nature (Kruitwagen et al., Vol. 598, 604-610). There is also a blog post by Lukas Kruitwagen himself on The Conversation.

Apart from the subject’s intrinsic interest, the study is noteworthy because it used machine learning (ML) to make the predictions. The base training dataset came from OpenStreetMap (although the paper makes a single mention and then, incorrectly, adds an “s”). The role of OSM is much better described in The Economist (paywall):

“For this, they turned to OpenStreetMap, an open-source rival to Google Maps in which volunteers had already tagged large numbers of solar plants. But there was little consistency. “Some people had just drawn rough outlines around an entire field,” Dr Kruitwagen says. “Others had gone in and traced the outline of each row of panels separately.” Fixing that involved a great deal of manual labour.”

There is also a New & Views piece in Nature as well. These are usually reserved for articles judged to have particular significance.

The Data

The data are available under a CC-BY 4.0 International licence at Zenodo: I downloaded the predictions and the training dataset, but, to date, have only looked at the former. Suffice to comment that there were a few slightly negative comments about data quality relating to the OSM source (see above). Note the 17000 items at 10 items an hour, which probably understates time spent searching, represent approximately 1 person year effort of data capture, and it is free!

Comparison

Comparison of solar power locations on OSM (blue) and those predicted by Kruitwagen (dark red). Original Image on Wiki.

I’ve done a couple of comparisons, but want to focus on China, because this is the area which may probably has the most large solar farms missing on OSM. I’ve looked a few times, but have generally found searching for solar by hand not very productive in China.

China

One very useful feature of this dataset is that it is richly annotated with a range of attributes, including country iso codes. It was therefore easy to select a subset for China. I did no filtering on size. There are around 18,500 of which about a fifth occupy more than 10 ha and half are estimated to have a capacity of 1 MW or more.

OSM data were pulled down using Overpass-turbo. Both power=plant and power=generator with a method tag of solar were needed as many large solar farms are still mapped as single generator areas. No attempt was made to exclude power=generator within power=plant, or very small household rooftop installations. I believe virtually all solar power mapped in China in the form of large scale solar farms.

I generated a simple intersections between the two data sets, and then found all items in the predicted data set which did not intersect an existing OSM object. Although the intersection finds over 7000 objects these correspond to about 2,500 items in the predicted set, leaving nearly 16,000 still to be found (see map above).

I’ve looked at a very small number and have not spotted any obvious false positives, although there are some structures which resemble solar panels which are difficult to interpret (possibly related to intensive agriculture). Often doing so highlights just how much mapping remains to be done in China, with whole villages still not marked in any way.

Europe

A little bit about comparison across Europe.

This also definitely identifies missing solar plants. In Britain where we know we have mapped pretty much every plant producing over 1 MW, a comparison picks out quite small scale generators of a few hundred modules (~100-250kW) which in general get picked during detail mapping of rooftop solar. Elsewhere the Oxford data picks out both large solar farms and significant rooftop installations on industrial buildings. Of the small number of locations I examined in Britain, a couple were false positives – both times polytunnels. Overall, the predictions look good and find really quite small sites.

Conclusions

This looks an enormously useful resource, which can potentially be used to help improve the coverage of solar power generation on OSM. The CC-BY licence may be a slight hindrance.

Note that OSM is already one of the best source of geolocated solar power data, see the press release of the Southampton group from last year, and our own work in the UK including rooftop solar published by Dan Stowell, Jack Kelly et al.. Let’s keep it that way!

Thanks to Dan Stowell who pointed me to the data & Jon Pennycook who noticed the article in The Economist.

A few weeks ago TrekClimbing asked on talk-gb about tagging various types of vegetation. He has documented some of the more problematic ones on the wiki.

The usage of natural=heath in the UK has not been particularly consistent, and his examples confirmed this. The situation is not helped because large swathes of natural=heath were added by a single mapper. Although doubts were expressed at the time, no-one was confident enough to say that such mapping was wrong.

Colourful late summer heath at South Stack

I have been aware that in upland areas of Great Britain natural=heath includes things I would describe differently. Notably these include : upland acid grassland used for rough grazing (mainly of sheep) and blanket bog.

It occurred to me that by comparing OSM data with another source would quantify this impression. Fortunately, a suitable source exists for Wales: Phase 1 habitat data.

Phase1

Phase 1 is a base level habitat survey for the United Kingdom introduced in the 1970s by the JNCC. It is at a finer scale (minimum feature size) than Corine and largely meant for ground surveying rather than remote classification. The main habitat classes and most of the detailed classes are understandable for laypeople. It also has an open access manua, which is quite readable. There are even convenient fold-out sheets of typical plant species for Phase 1 survey of Heaths.

I maintain, in a somewhat desultory manner, a wiki page showing correspondences between OSM tags and a subset of Phase 1 codes. Phase 1 codes can be tagged directly using the plant_community key.

Wales was unusual in that the official nature conservation body at the time CCW now NRW) decided to survey the whole country. The Welsh data has been available since 2005, more recently under an Open Government licence. A number of counties in England also carried out area-wide surveys, but these are not public.

Upland grasslands seen from Moel Eilio.

Analysis

I extracted all areas tagged with landuse or natural for Wales transforming them to the British National Grid. Although most natural tags represent habitat classes, a few (peninsula, mountain_range) represent geographical concepts . These latter were excluded from further analysis.

The basic steps were to find all OSM polygons which intersected with the Phase1 heath categories ( prefixed with “D”), and the converse, all Phase1 polygons intersected with natural=heath. OSM tags were grouped into broader categories (woodland for natural=wood and landuse=forest; grassland for natural=grassland, natural=grass, landuse=meadow etc). The results are shown below:

Fortunately pretty much all Phase1 heath polygons correspond with natural=heath (or natural=fell, which may by now be a synonym), where any type of habitat or landuse has been mapped on OSM. As expected, the reciprocal relationship covers a range.of habitat classes. The most significant, accounting for 90% of the area were:

• Grassland (Phase1 “B” classes)
• Heath (Phase1 “D” classes)
• Bracken (Phase1 “C.1.1” in the tall herb “C” group)
• Wetlands (Phase1 “E” classes, most being E.1.6 and E.1.7 : blanket bog.

OSM heath compared with Phase 1 codes

The last step was to take the most common Phase 1 classes (“B”,”C”,”D” and “E”) and find all OSM polygons overlapping these. I’ve chosen to represent these with an alluvial flow diagram (apologies for it’s somewhat rough and ready appearance: I haven’t mastered the available R packages (alluvial & ggalluvial) to allow finer control of elements).

Flows between detailed Phase 1 codes, OSM tags and Phase 1 generic classes. The original diagram is here, and an alternative (ggalluvial) version here.

A direct comparison with these Phase 1 codes shows a close correspondence of natural=heath and the relevant Phase 1 polygons.

Caveats

This analysis did not perform any extensive processing on either dataset. The following issues might affect the results:

• Invalid polygons. The Phase1 data has a number of invalid polygons. I merely used ST_MakeValid, and excluded any not do repaired. Phase1 data was compiled some time ago and habitsts have changed (most obviously with forestry plantations)
• Overlapping OSM polygons. Despite the desire of some for each area of land to have a single landuse or natural polygon, there will always be overlaps (as natural=peninsula shows). For upland heathland typical overlaps are water bodies, and conifer plantations. These will have resulted in some overcounting although I believe the effect is very minor.
• OSM tagging. Some polygons have both natural & landuse tags (e.g., natural=heath, landuse=military). A few of these are inconsistent (natural=wood, landuse=grass). I arbitrarily selected one of these values. A few typos were corrected but a polygon tagged natural=health had been “corrected” so many times I could not reliably restore the original mapper’s intent (it is on a remediated colliery spoil heap adjacent to a park: similar places exist in Notts and are part of Country Parks).
• Mosaic Phase 1 polygons. These were not included in the analysis.

Findings & Conclusions

The main finding is that various Phase1 habitat codes have a high correspondence with natural=heath. This is not true for some other codes, such as “B.4” grassland. This is a positive result, in that natural=heath usage seems to represent a coherent range of, predominantly upland habitats. I would expect similar patterns hold on much of Europe because the Corine code mapped to natural=heath includes Moorland.

I know this does not really answer the original queries, but by looking how a tag is actually used on OSM I think it is possible to identify some ways to resolve some of the issues.

Recommendations & Next Steps

I think I can formulate a simple set of recommendations & follow-up research:

• Use a heath subtag. Given that natural=heath appears to represent a coherent group of habitats in Wales, and elsewhere in the British Isles, the obvious way to map the clearly distinct habitats represented by the tag would be through use of a heath subtag, for instance heath=blanket_bog or heath=bracken. There is already some usage on OSM, although dominated by the undocumented ‘heath=alvar’ (all on Gotland).

• Check lowland heaths. Wales is lacking much in the way of lowland heaths which are common in Southern England, and have seem to closely match expectations. So any implications for such heathland need to be considered.

• Check natural=heath elsewhere in Europe. Much natural=heath will be dominated by Corine imports. Again I have the impression that this results in heath having a broader meaning on OSM than its classic ecological interpretation.

• Compare Corine data for Wales. A corollary of the above is to make the same comparison for Wales but using Corine data.

• Map missing areas of heath in Wales identified from Phase 1.

• Document the sub-types of heath mentioned.

This latter item is what I plan to do next.

I was intrigued by the photo on the wiki illustrating the hardest level of the Swiss Alpine Club scale for mountain hiking (known as SAC Scale on OSM and tagged with sac_scale). I couldn’t identify the location and wondered by looking for ways with the tag I could do better.

A quick overpass query revealed widespread misuse of the sac_scale tag for true alpine climbing routes which the Swiss Alpine Club grades using a completely different scale.

A few of the more egregious examples:

• Geant-Rochfort Arete. A classic high-level snow ridge encompassing a couple of 4000ers. Tagged with highway=path and sac_scale:

Description of route on Hikr.

• Biancograt of Piz Bernina: apparently trail visibility is horrible, which is a bit of a surprise, because the ridge is rather narrow.

• Normal route for Piz Palu. At least this is what it looks like to me, although it is labelled “Piz Spinas - Cresta SW” and when I’ve been at Diavolezza the route through the ice falls has been somewhat different. Although this route is immensely popular most of it is over a very steep glacier with two major ice falls. The existence of significant objective dangers was used in the 1929 movie Die weiße Hölle vom Piz Palü starring Leni Riefenstahl.

The normal route from Diavolezza contours above the crevasse field in the foreground and then passes through the lower ice fall quite close to the rock wall on the left before ascending to the minor col to left of the E summit.

There are numerous other normal ascent routes for significant peaks in the Pennine Alps. It is obvious that the tag has been used as follows:

• normal ascent routes which will be heavily used in the alpine climbing season. In good weather the route taken by climbing parties will be clear, but once fresh snow falls earlier traces are often effaced. On rock it is still possible to lose the line of a frequently climbed route.
• as a proxy for accurate Alpine Climbing grades (F/PD/AD/D/TD/ED or L/WS/ZS/S/SS/AS etc)
• often with a highway=path tag (although this was deleted in the case of the Matterhorn)
• sometimes with surface or trail_visibility tags to perhaps indicate these are not actually paths.

As normal hikers are extremely unlikely to undertake a difficult_alpine_hiking tour (T6 grade), the immediate impact on OSM users is probably not anything to worry about. In the past I followed a few T5 routes, and they were slightly above my comfort level, so I would never have undertaken a correctly graded T6 on my own. However the difference between T6 and all these routes is much greater than that between T5 and T6. They require much more equipment and very different techniques: safe glacier travel, crevasse rescue, climbing at UIAA III, etc).

Elsewhere in the world there have been problems with people assuming a path on OSM represents something within their capabilities. We have a responsibility to ensure the data within OSM does not exacerbate such issues.

The other day a new user was asking questions about mapping the new development of the former Royal Engineers Barracks opposite Mill Hill East tube station. I’ll write about their problems with postcodes another time.

However, this prompted me to look at the OSM map of the area, which I knew well in the 1980s. I was a PhD student in Central London (UCL), but a close friend was doing his doctorate at (NIMR](National Institute for Medical Research), and I visited fairly frequently. Over time I got to know other scientists & staff at NIMR, not least because I became an active member of the walking group which was part of the staff club (NIMROD, which still leaves a trace on OSM), but also because of scientific collaborations.

NIMR has been subsumed into the new Crick Institute behind the British Library, and the imposing buildings have been knocked down & are now being re-developed for housing. I presume fairly high-end housing because the location is on a ridge with wide views across NW London and still some remnants of farmland in the small valley to the NE. Three roads have been named:

• Medawar Drive: Sir Peter Medawar was a Nobel laureate in Physiology, and director of NIMR until he had a stroke, when he move his research lab to another MRC institute, the CRC at Northwick Park. A friend was his chief technician, and he collaborated extensively with the head of my unit.
• Cornforth Lane: Sir John Cornforth was another Nobel laureate (Chemistry 1975), who also spent part of his career at NIMR
• Rosalind Close: we see the pattern, distinguished scientists associated with the institute. So why a female given name, rather than another surname? Well, there is one very well known woman scientist with this given name Rosalind Franklin, who if she had lived certainly deserved the Nobel Prize. I don’t think she had any direct association with Mill Hill having spent her career in France and then at King’s and Birkbeck Colleges.

Of course there may be an innocent explanation for these choices, but on the surface the above seems the only reasonable interpretation. It’s a shame the developers could not have marked some of the remarkable women scientists who worked at Mill Hill, people like Brigitte Askonas or Rosa Beddington.

Another short post on street furniture (you can definitely tell I haven’t got out much in the pandemic as this is the sort of mundane detail I’m focused on at present). Once again these are, as much as anything, notes for myself.

The Ranty Highwayman wrote a great detailed post about the different types of tactile paving and how they should be used in the UK: https://therantyhighwayman.blogspot.com/2015/02/tricky-tactiles.html. It’s very helpful if confused about the various types, and may help to spot incorrect layouts (which can be tagged).

My only additional contribution is to note that in the UK the tactile paving slabs tend to come in a very restricted set of sizes:

• 40 cm square. The most usual size for both blister & corduroy (hazard warning) paving. There are 6 blisters up & across. I haven’t counted the lines.
• 45 cm square. A bit less frequent, 7 blisters across (illustrated in the blog post above).
• larger (perhaps 60cm+). Blister interval possibly bigger than in the previous two examples, 9 blisters a side. I haven’t measure these slabs precisely, but this is an example on Mapillary.

More examples of tactile paving can be found on suppliers’ sites such as this one.

So far, so unbelievably nerdish, but there’s a nifty bit of knowledge from this info which can be applied to mapping much more widely! Knowing the paving slab sizes makes it easy to estimate the width of pedestrian crossings, crossings at dropped kerbs and any pavement/sidewalk where the corduroy hazard slabs are used. This can not only provide a simple way to survey widths without using a tape measure or laser device, but can also assist in estimating widths based on experience.

Over time I’ve noticed features appearing on highways & paths which are unusual and which I don’t know how to describe. Often these become much more prevalent. Recently I made an effort to do some reasearch on a couple of these and learnt some useful terminology (documented here so I don’t forget):

• Deterrent Paving: These is a form of paving which it is difficult to walk or drive over. It is used in a range of situations, but a common use is where a road has been blocked for cars and cycles are allowed through but there has to be some provision for access by emergency services (often labelled “FIRE PATH” in the UK).

There is quite a wide range of different types, some aimed more at stopping pedestrians. The presence of deterrent paving may represent a significant hazard for people with reduced mobility or impaired sight.

• Perforated Kerbs: These are kerbstones which also act as part of the road drainage system. They are being installed quite extensively local to me. The kerbs have numerous holes and are hollow: they link up to form a drainage channel. Obviously they have the potential to remove surface water rather more quickly than traditional road gullys (storm drains. They are also used in more elaborate Sustainable Urban Drainage Systems (SUDS). I found this brochure gave a good overview.

At this point I have no real thoughts as to appropriate tags for these things, but felt it worthwhile drawing others attention to them.

In March we celebrated the 10th Anniversary of our first pub meeting of local mappers in Nottingham. I’ve written a bit about on-line and in-person meetings held this year, but much more a set of reflections about how the group and its meetings have panned out over the past 10 years.

I hope there are some nuggets of value to others either already running such meetings or contemplating doing so.

Lastly, a big thank you for everyone who has come along over the years, and to Kyle who prompted me to write much of this in the first place.