Using OpenStreetMap to map urban pedestrian facilities

Most of the maps on this site were created at least in part with GIS data from OpenStreetMap (OSM). OSM (for those who do not know) consists of GIS data—that is geographic data in a format usable by computer mapping programs—for the entire world. It differs from many other sources of GIS data in that its files are available to anyone for free—and are supplied in part by local contributors, who use information from aerial photos, satellite images, and personal knowledge. (OSM also uses publicly available government datasets.) OSM GIS data, in general, seem to be reasonably good, as several scholarly papers have testified.¹

One of OSM’s contributions to the process of amalgamating data from many sources has been to standardize it. This activity isn’t simple, and it inevitably involves some arbitrary decisions. An example of this can be found in the files for railroads. Railroad files are supposed to be track maps whenever that’s possible.² A railroad consisting of four parallel tracks will be portrayed by four parallel lines. Not only is every track shown; so are switches, marshalling yards, and repair facilities. These extra tracks aren’t very visible in maps at a small scale (that is, maps that show large areas), but they do appear in highly detailed urban maps. Because my focus in most of the maps on this site is passenger rail routes rather than tracks, I’ve often edited the extra tracks out (see the maps of Tokyo railroads for example). This is a time-consuming process, but it does result in cleaner maps. It’s doable in part because OSM data are reasonably standardized, and editing can be done in a somewhat mechanical way,

A major component of OSM’s standardization work is its classification of geographic features. Roads, for example, are classified into more than a dozen different types that are the same for the entire world. There are inevitably some issues of consistency. The category “service road,” for example,  is the lowest-ranked type of road for motor vehicles everywhere. It’s used for alleys in American Midwestern cities, and for access roads in the massive blocks of “new towns” in Hong Kong and Singapore. These aren’t quite the same thing, but neither are the urban landscapes being mapped. There really isn’t much to be done about this problem except to be aware of it. In general, one can count on OSM road data to be reasonably consistent for different parts of the world. The same is true of railroad data, where classifications like “rail” (for mainline railroads), “light rail,” “subway,” and “monorail” are used everywhere. Land-use data are similarly pretty consistent: the world’s parks may not all be the same, but it’s usually fairly clear what’s a park and what isn’t. One can count on data for water bodies to be reasonably reliable too.

“Pedestrian features,” however, are another story. The classifications are wildly inconsistent, perhaps because these have not been quite as important to OSM’s editors as, say, roads and railroads, or (arguably) several other feature types.

“Pedestrian features” are always treated as a special type of road. There appear to be four main types and several additional, less common types. The four main types are “footways,” “paths,” “cycleways,” and a type labeled simply “pedestrian.” Less common types of pedestrian features are “steps,” “tracks,” and perhaps “bridleways.”

Let’s take a look at how these terms work in practice. I’ve included below maps of the central parts of several cities.³ All these maps would be at a scale of 1:50,000 if printed on 8.5 x 11 inch paper. They use the same color scheme I’ve been using in recent posts. Red = subways or other rail rapid transit lines. Orange = streetcars or light-rail lines. Brown = other passenger rail lines. Grey = roads for cars, usually excluding “service roads.” Light blue = water. Light green = parks. Dark green = pedestrian facilities, as specified. Medium green = steps on the Lyon and Moscow maps.

I’ve included maps for central Chicago, Lyon, Moscow, Kyoto, Hong Kong, and Dubai. My sense is that adding additional cities would have suggested that the problem is even messier than it appears to be from the examples given here.

Here are the maps. Commentary follows.

“Footways” are probably the least consistently applied classification. “Footways” are supposed to be “designated footpaths; i.e., [paths] mainly/exclusively for pedestrians. This includes walking tracks and gravel paths.” But in Moscow every sidewalk is classified as a footway, and in Chicago every sidewalk in parts of the Loop and the Near North Side—but not elsewhere in the city—also gets this classification. This makes absolutely no sense to me, if only as sidewalks are essentially universal in central Moscow and Chicago; they are present along every road that isn’t a limited-access highway. Somewhat similarly,  in Kyoto most sidewalks along major streets are considered to be footways. Because sidewalks are often not present along minor roads in Kyoto (and other Japanese cities) and can be very narrow even along some important streets, it’s possible that an attempt is being made to note particularly wide sidewalks, but this process seems pretty arbitrary to me. In Lyon, in contrast, footways include a more modest (and interesting) set of features: pedestrian walkways along the Saône and Rhône; paths in parks; pedestrianized streets; and pedestrian passages in the hills on the West Bank of the Saône. In Hong Kong, the designation footway is used for many things but never ordinary sidewalks: pedestrian paths along waterfronts; pedestrian paths in parks; walking paths in the hills; the elevated walkways that thread through parts of Central; and bridges across major roads, which are quite common. This covers a lot of territory. Similarly, in Dubai, footways include some of the pedestrian paths along Dubai Creek; paths in parks; and Deira’s pedestrianized streets. Sidewalks seem to be excluded.

Employment of the classification “pedestrian” seems a little more consistent. It’s supposed to be used “for roads used mainly/exclusively for pedestrians in shopping and some residential areas which may allow access by motorised vehicles only for very limited periods of the day.” This could be interpreted to mean urban streets that have been pedestrianized, i.e., generally closed to motor vehicles. This category does seem to be used this way in Moscow, where several east-west roads off Tverskai͡a Street north of Red Square as well as in the Zamoskvorech’e District are pretty classic examples of pedestrianized streets. Some pedestrianized streets in Kyoto’s main shopping district and in Vieux Lyon are also classified this way as is the pedestrian corridor that replaced Ogden Avenue in Lincoln Park, Chicago. But “pedestrian” is also used for parts of the walkways along the Saône, the Moscow River, and Dubai Creek, which have a very different origin. And, as noted above, in some cities, pedestrianized streets are classed as footways.

The much less common classification “path” has analogous problems. It’s supposed to be used for “non-specific paths,” that is, those not exclusively for walkers, cyclists, or horse riders. This could, of course, apply quite widely, but in practice “path” is used sparingly. On the maps above it’s used most commonly on the Kyoto map, where, so far as I can see, it overlaps categories like “footway” and “cycleway” in ways that I find hard to understand.

“Cycleways” are bicycle paths. This sounds easier than it is. The OSM guidelines are concerned as they should be with whether cycleways also permit pedestrians but seem not to suggest a completely consistent solution. Also, it’s not clear whether lanes painted along streets should be included. They aren’t on the Chicago map; even protected lanes are usually excluded (and the western route in Lincoln Park that is included is not actually a bicycle path). In contrast, on the Moscow map, the narrow, little-used (and perhaps somewhat dangerous) lanes that follow the Boulevard Ring are considered to be cycleways, as are several bike paths painted on sidewalks. The Lyon map shows both separate paths and protected lanes as cycleways. There are no cycleways in central Hong Kong, so I’ve used the fourth Hong Kong map only for steps.

“Steps” are another type of pedestrian facility. They are common only in hilly cities, and I’ve included them only on the Lyon, Moscow, and Hong Kong maps. Unlike the other categories here, “steps” seem to be applied quite consistently.

“Tracks” are supposed to be used for “roads for mostly agricultural or forestry uses.” On the maps above, the category is used only in Hong Kong, for some of the unpaved trails in the mountains, which makes some sense, I guess (although the distinction between tracks and paths is inconsistent). I haven’t included tracks on any of the maps above.

The chief problem of consistency occurs for the categories “footways,” “pedestrian,” and “paths.” They are definitely not distinguished meaningfully. It might make sense to merge these categories, especially if sidewalks could be eliminated from the “footway” category.

Bridges over and tunnels under roads are another problem. Bridges over roads are noted assiduously in Hong Kong and Dubai, where they are common. The chief problem is cartographic. If you include these in a map at medium scale, you get a lot of odd-seeming green smudges on the map.  Tunnels are extremely common in Moscow (and other formerly Soviet cities), but they don’t seem to be noted in the database. Perhaps that’s just as well, as there is no very nice way to map these except at the largest scale.

In maps on this blog, I’ve typically edited the data for pedestrian facilities quite a lot. For example, I’ve generally taken sidewalks out of “footways.” This has worked pretty well, I think. But it would perhaps be better if less editing were required, and if the categories recognized in the database actually corresponded in a consistent way to differences in the real world. For pedestrian facilities, they really don’t.⁴