As I’m interested in pedestrian infrastructure, I asked myself what percentage of all sidewalks in Germany has already been mapped.

It is difficult to answer this question because I do not know the total number of sidewalks in Germany. However, it is possible to change the question slightly to make it easier to answer: How many streets in Germany contain information about the presence of sidewalks?

This is easier to answer, because we can assume that all streets of Germany have already been mapped. Therefore, to answer the question, we just have to count the streets that contain the desired information.

 

Sidewalk tags

Currently, four tags are in use for mapping sidewalks: sidewalk, sidewalk:left, sidewalk:right and sidewalk:both. The tag sidewalk was introduced around 2010 and is still widespread, accounting for around 80% of all sidewalk tags:

The other three tags were added in January 2021. Normally sidewalk:left and sidewalk:right should be used together to provide a complete view of both sides of the street. The tag sidewalk:both is just a shortcut when both tags have the same value.¹

Speaking of values: These three tags can take the values no, yes and separate, meaning in order: There is no sidewalk, there is a sidewalk, and there is a sidewalk which is mapped as a separate highway.

In addition to these three values, the sidewalk tag can take the values both, right, left and a deprecated synonym of no called none. Only the values both and no privide a clear picture of the intended meaning when mapping.

The values separate and yes do not distinguish between sidewalks on both sides and a sidewalk on one side without specifing the side.

Likewise, the values right and left are ambiguous: They obviously state something about the sidewalk on one side, but they may or may not imply what is on the other side of the street.

To make sense of this, I started thinking in terms of sidewalk:left and sidewalk:right. Unlike the other two tags, these give information on exactly one real object. If the sidewalk tag is given, mapping of both and no to sidewalk:left and sidewalk:right is straightforward.

If the sidewalk has one of the values left and right, I use them to define one of the two tags, leaving the other one undefined.

If the value is separate or yes, I’m using a new value, maybe, for the sake of counting.

Finally, I consider the data invalid whenever a sidewalk or sidewalk:both tag is present alongside any of the other tags. (Fortunately this happens very seldom.)

 

The Art of Counting

I ended up with the following diagram, using this. (red: yes, green: no, yellow: separate, cyan: maybe, purple: invalid)

More than 90% are missing? I expected a large number, but this is unbelievable. On second thought I realised I had made a mistake: I counted the information on sidewalks for all highways, but this cannot be expected for many of them, such as https://wiki.openstreetmap.org/wiki/Tag:highway=footway.

So I recounted, this time only considering highways with value primary, primary_link, secondary, secondary_link, tertiary, tertiary_link, residential and unclassified. This time I got the following:

This looks a little bit better, but it still isn’t, what I was looking for. After thinking about it some more, I realised that I’m actually interested in the number of sidewalks in urban areas. Here I counted everything outside of town too.

Unfortunately, OSM does not provide direct information on urban areas.² But luckily, I wrote a rather complicated program a few years ago to estimate these areas using other OSM data. Reusing the results of this program, I created the following diagram:

It’s not much better. About 2/3 are still not mapped. I made another mistake, though: I counted OSM objects instead of street kilometres. This means that a very short piece of a street counts the same as a very long one. I therefore ran the program again, this time measuring the lengths of the ways, and I got this final diagram:

The result is worse than what I got before: 3/4 of all urban sidewalks have not been mapped yet!

 

History of the Sidewalk Tag

Is the mapping of sidewalks stalled? Or did it just begin recently? I took some OSM data snapshots from the past to get some idea how the mapping of sidewalks developed over the years. This is the result:

It’s clearly visible that the introduction of the three new sidewalk tags in 2021 caused an increase in sidewalk mapping, but this increase already starts to decline again, long before we got near 100%.

 

Missing Feedback

Well, the question now is, why are still so few sidewalks mapped? I’ve got a theory which, I think, is quite convincing: Mappers don’t receive any feedback when they map sidewalks.³ OSM Carto (the main map) doesn’t show this information, nor any of the other maps featured on the OSM main page.

Also, pedestrian routers do not rely on these tags, because so many are missing; they also route on streets without them. Thus one doesn’t notice if the tag is there or not. It’s a vicious circle.

What we need, is some feedback. Ideally, OSM Carto would show this information, but I fear, this will not happen. The goal of this map isn’t to show all mapped features, but to provide an all-purpose map.

A pedestrian map on the main page would also be great, because there isn’t one, and it would make sense to show the sideway tags on such a map. However, I’m not aware of any effort to generate such a map. It might take too long until this happens.

So, at least those of us, who want to add this information need something else. As far as I know, Overpass Turbo is updated in realtime, which makes it a perfect instrument for immediate feedback. Thus I created a query: Overpass Turbo query showing missing sidewalks. You can use this query on your favourite places to see where sidewalks have not been (fully) mapped yet.

Here an example showing the area around Bad Godesberg in Bonn:

 

Some Additional Pointers

• I published the program I used for counting on GitHub.

• I started writing this blog post about a year ago, but for unknown reasons, I left it unpublished until now. In the meantime, Dr. Roland Olbricht gave a talk on this subject on FOSSGIS 2025 conference (in German). He too realised that information on sidewalks is often missing. However, I do not agree with all of his conclusions, particularly his view that the value separate should no longer be used.⁴

 

---

1. I dislike the use of sidewalk:both, because I strongly prefer 1:1 mappings between real objects and OSM data. The tag sidewalk:both is a 2:1 mapping that increases the workload for consumers of the data. However, it is still much better than using the problematic sidewalk tag. ↩

2. Actually, the idea of “urban areas” is a mistake in itself. Imagine two streets crossing with a bridge. It is possible for one to be urban and the other not. It would be better to decide for each street separately if it is urban or not. I think, this approach is also not easy to program and I haven’t found the time to give it a try yet. Thus I’m sticking with urban areas - it’s better than nothing. ↩

3. If you map sidewalks as separate footways they will appear in OSM Carto. However, the corresponding sidewalk tag may be missing or incorrect. ↩

4. In his talk, Roland gives an example of a mapper misunderstanding the meaning of https://wiki.openstreetmap.org/wiki/Tag:sidewalk=separate, and states that there are more such examples, rendering the value useless. He recommends using https://wiki.openstreetmap.org/wiki/Tag:foot=use_sidewalk instead. Unfortunately with https://wiki.openstreetmap.org/wiki/Tag:foot=use_sidewalk alone, you still do not know, where the sidewalk is nor whether it has been mapped separately or not. (But of course, as an addition https://wiki.openstreetmap.org/wiki/Tag:foot=use_sidewalk is fine.) However, this information is important for real pedestrian routing (with “real” I mean routing on the sidewalks instead of routing in the middle of the street). In my opinion, the meaning of the value separate is quite clear: there is a sidewalk and it has been mapped separately. In the counter-example, the mapper used this value for a footway that ran parallel to the street but was not a sidewalk. ↩