I’ve recently been attempting to edit in Sheffield City Centre, within the ring road, and I’m having an awful time whenever I do. This is due to what I’ve begun terming “mapping debt” after the concept of technical debt in software engineering (IDK if I’m the first to do this or if it has another name.) Nothing is really broken about the city centre mapping, it’s functional, but it’s an absolute nightmare to add to, edit or improve.
I can currently think of the following problems with the mapping in the city centre:
Alignment issues: Multiple people have seemingly mapped at varying alignments over the years with little regard for the alignment of the surrounding data. A lot of the older edits seem to be misaligned by about 3-5 meters and newer edits seem to have been mapped in line with Bing imagery which seems to be pretty good in terms of alignment, within a meter in most places. This means that in adding features, especially separate footway=sidewalk ways I end up having to correct the alignment of buildings or other features as the alignments are all different. So far I’ve been correcting them to Bing given it’s close enough and the cadastral parcels are not as useful in the city centre.
Gluing: Lots of the city is glued together including lots of landuse=* areas that are glued to roads. I have largely left these alone whenever possible but every now and again it becomes incredibly frustrating to deal with especially when dealing with the next issue.
Inaccurate/Oversized Buildings: A lot of the buildings are oversized, have inaccurate geometry or are aligned to the rooftop rather than the foot of the building (though it’s hard to know what something was aligned to given the aforementioned alignment issues.)
Misc Warnings: There are so many warnings in the city centre that I end up having to wade through a mess of them every time I make an edit, I counted about 10 per street I add separate footway=sidewalk ways to, this then snowballs as those features then have multiple issues themselves and resolving the warnings often takes longer than the changeset initially took.
Given all this debt I think it would be a good idea to work out a system for cleaning up the city centre to make future mapping easier. I’m looking for suggestions on how to focus effort to unpick the mess as it’s all linked together into one giant tangle. Is anyone aware of any other large scale city centre clean-up efforts, especially in areas with poor quality imagery (Bing is really not great resolution here.)
I think this illustrates a delicate balance: Adding new (and presumably (or at least hopefully) correctly assigned items), vs aligning items in a potentially unlimited area to these new items. I am not sure there is a perfect answer to this - or at least, I have not found one. I take the approach of trying to best align with a definitive dataset, but then align local geometry as far as possible at the edges.
I believe, anecdotally, that the consensus is to unglue these - which should not result in any reduction in data fidelity.
I suspect these will need to be realigned manually, which should be a very helpful process for OSM in itself. There may be future options to align these using imagery and cadastres automatically, but from experience the areas in the centre of cities and towns are adverse to this approach as there are too many ambiguities.
As unfulfilling as it may sound, I’d personally suggest ignoring these unless you have edited the related items, or are keen to fix them. Even then, if you can’t solve the problem, and it doesn’t impact on your edits, I would advocate for not feeling obliged to fix it. They will still be there for someone else to look at, and don’t diminish your improvements.
This is only a small subset, based on my personal perspective. In the majority of cases, the normal approaches apply, but I have personally found there may be some queries around the mapping of:
Shops with residential accommodation above
Areas above shops that span several shops - where a shop has been subdivided or in cases where there are “overhangs” another building (for example in Lincoln).
The problem with this approach is that when submitting an upload in JOSM I then need to check through all the warnings to make sure they aren’t issues introduced by my changes. Given I edit a few roads at a time it becomes pretty taxing to understand which ones are related to my changes. I get that we can’t make the city centre warning free but currently the number of warnings interfere with my willingness to edit because of how cognitively taxing the warning make it. It’s almost easier to fix everything than it is for me to ignore them and keep track of which warnings are related to my edits.
Thank you for these suggestions, I’m especially interested in things that could be loaded into the JOSM todo plugin and fixed in semi-isolation so these are of interest.
That would be a massive help! I’m wondering if it’d be a good idea to break up the landuse like the commercial and the residential into the blocks of buildings? Get them off the roads altogether so they’re smaller and easier to work with while also reducing the amount of clutter people focusing on the roads have to deal with.
Lemme know what you think on this. I’m much less skilled with landuse shenanigans but will also give a go at detangling.
In areas which have a lot of older buildings and street alignments which haven’t changed, I’ve found that using the NLS layers with out of copyright 1:1250/1:2500 OS maps can be a very useful tool to start the realignment process in JOSM. You’ll need to adjust the alignment of the NLS layer to match INSPIRE as it’s often very slightly out and can change over small distances.
Thanks for raising this. It’s too much for a single mapper to bear, I think, so it may be worth dividing the city centre up into “cake slices” and giving them names or numbers. Set a topic such as landuse ungluing, then people can then pick from a list of areas to work on without worrying about undoing or duplicating the efforts of others. Visible progress should spur the effort along. Then repeat for building alignments, pavements and footpaths, whatever needs attention…
What kind of divisions are you thinking? Also what would be a good tool to organize that effort? I don’t particularly want to start a HOT project and attract careless mapping for the sake of completing a task.
I was thinking we split the city centre up by city block and fix the landuse and buildings in each one and then split the city up based on road and work through each road and it’s associated features. Could be a MapRoulette task or something similar with a specific set of guidelines to get the city up to a consistent standard?
It depends. Someone might only want to add pavements & crossings, and would record all the sectors where that task has been completed. You only need to claim a sector is there is a risk of conflicts which depends on how many people participate and how targeted the mapping is.
There are many large landuse polygons that span the entire centre, these should be split up into smaller ones before any work is divvied up. However I think small alignment changesets are probably better and will have little chance of conflict.
I see others have replied before I had a chance to
I had an idea of dividing up the city like @LordGarySugar has laid out, but with smaller divisions: something that could be covered on a lunch break. (Insert joke about Division Street here.) Looks like SimpleTaskManager is the way to go, per @ceirios. Also, perhaps slightly against the spirit of openness but a self-selecting team would be best for the tasks rather than throw it wide open to anyone. There seems to be enough interest on this thread already.
FWIW you can define SimpleTaskManager task areas in other software and import them. I think they can be any shape you like then.
I’ve mostly used this when I did want squares but I wanted to only create them within e.g. 50m of a road in a sparsely built area.
I do wish it supported on the fly divisions like the HOT Tasking manager and Divide and Map Now but It’s still my preferred site for personal projects.