User:Ftrebien/Drafts/Generic highway classification principles

Main motivations for this: tackling lack of objectivity, enabling mappers to reach agreement with the fewest possible uncertainties, enabling mappers to map abroad with fewer mistakes, facilitating exchange between countries, reducing user parsing overhead by making maps look more similar across different countries.

Inspirations

• Proposed features/Highway key voting importance
• Netherlands roads tagging
• American functional classification system (dead link, see web archive) (adopted by Waze)
  • Deleted without discussion while there's manifest interest by some mappers
  • Functional classification criteria
• European functional classification guidelines
• Dutch functional classification and cycleway design
• Canadian tagging guidelines
• Australian tagging guidelines
• Attribution of roads in Germany (translated)
• HOT's highway tag usage in Africa
• Opinions/perceptions on displaying unsealed major ways (partially dismisses highway structure/surface as criterion for classification)
• Permeability: useful concept for classifying tertiaries in organic (irregular) urban areas
• Comparison of existing 2, 3, 4, 5 and 6 class systems (and others); key observations:
  • "The topics which RH focus on appear to be, essentially, access and speed."
  • "The distance element of access (how many junctions or access points per km) should be based on the concept of road network mesh."
  • "Terminology – different authors writing at different decades use the same term but not necessarily with the same precise meaning. Imprecise use of terminology for a desirably precise technology such as highway engineering is a bit dodgy."
  • "It seems to me that an air of respectability surrounds a concept (RH / FC) which – although it might be useful – nevertheless uses vague terminology, is not well-quantified, ignores practicalities from the real world and is described differently by different writers." My note: however, for OSM's apps, classification is important, such as deciding which roads to render or which roads to consider first on a routing algorithm.
• Road hierarchies; key observation:
  • "In the real world many different organisations use a region’s road network. These organisations have their own tasks and priorities, and so can each give a different level of ranking to one and the same road. It is even possible that one road may be very important for one organisation but completely un-important for another. This argument is not merely a play with words. Today’s road network managers have to provide a safe, functioning road network which matches the needs and priorities of the road users – which means that half their task is understanding what the route needs and priorities of these users are."
• Possible classifiers: shows how complex this discussion can become and why we should stick to simple rules that fundamentally capture a generic sense of importance
• Are motorways necessary?: a statement on inverted notion of importance, food for thought
• Residential Street Pattern Design
• (functional) highway classification is sometimes partially related to urban planning and/or (functional) zoning
• Grid plan section Ancient grid plans:
  • "In the 1960s, traffic engineers and urban planners abandoned the grid virtually wholesale in favor of a "street hierarchy". This is a thoroughly "asymmetric" street arrangement in which a residential subdivision—often surrounded by a noise wall or a security gate—is completely separated from the road network except for one or two connections to arterial roads. In a way, this is a return to medieval styles: as noted in Spiro Kostof's seminal history of urban design, The City Shaped, there is a strong resemblance between the street arrangements of modern American suburbs and those of medieval Arab and Moorish cities. In each case, the community unit at hand—the clan or extended family in the Muslim world, the economically homogeneous subdivision in modern suburbia—isolates itself from the larger urban scene by using dead ends and culs-de-sac."
• City street orientations around the world with examples of oriented grids and its variations and also examples of more organic systems
  • Organic cities in North America: Charlotte, Pittsburgh, Boston
  • Regular grids outside North America: Beijing, Barcelona, Buenos Aires
  • Grids with multiple orientations: Detroit, Denver, Shanghai, Amsterdam, Turin, La Plata
  • Paper: "In many cities, planned and spontaneous spatial patterns co-exist, as the urban form evolves over time or as a city expands to accrete new heterogeneous urban forms through annexation and synoecism."
  • Comments:
    • "Many cities with medieval origins have essentially a polar coordinate system with a central square (often at a church) and some roads in the 'spoke' direction and others in the 'circumference' direction. An analysis of cities based on Cartesian coordinates will not do the regular order of these cities justice."
    • "This is certainly the case for Paris, which is perfectly 'legible' when understood as wheels and spokes divided into inner and outer slices (arrondissement)."
    • "It would be interesting to have direction of major geographical feature on the plots. Clearly New York and Barcelona are based on the sea shore. While Budapest and Warsaw on the main river."
    • "Cairo is only vaguely aligned along the cardinal directions because of the Nile."
    • "I suggest for a fun American city try New Orleans. The lack of cardinality here is no notorious that no one uses the cardinal directions. The curvature of the river forces a lot of changes in the grid orientation of the city."
    • "Can we assume that some cities are originally designed for cars (driving from one place to another easily) and some cities are not? For example, Atlanta versus Rome or Denver versus Paris?"
    • "Boston makes a whole lot more sense if you go back and look at a pre-in-fill map. There are multiple different grids based on the wide points in the land, that neck down to pass through isthmii. Then the land around an isthmus got filled in, and the new development followed the new coast line. Then repeat."
    • "Toronto streets are roughly parallel or perpendicular to the Lake Ontario coastline. This is generally the case throughout peninsular Ontario. In the centre of the peninsula (Kitchener / Waterloo area) the township grids come together in a muddle."
• It is possible to draw an inspiring analogy between highway hierarchy and natural branching structures, such as the venation of water lily leaves (more geometric and almost orthogonal near the border, meant for structural support) or this or this (also geometric, but a circulatory system) or gorgonian sea fans (more free, also a circulatory system): those are reasonably planar networks, grow from a central region, with veins of clearly different capacity, with main veins at the tips becoming as small as "secondary" veins near the network core (if you look closely there is also a "tertiary" vein network branching off inside each cell). (Actually, there is distribution science behind this analogy.)
• are there other important sources to consider?

Core ideas

Using place=* as a marker of importance. In many European countries, that key is associated with political/administrative status. This is interesting because it brings together many potentially conflicting interests (e.g.: concentrated wealth vs population size vs niche activity such as military bases or tourist destinations). As such, it is a factor that helps OSM maintain itself neutral to diverging interests. Where such an association hasn't yet been made (many of the countries with a less developed OSM community), it is associated with population size, which is an easy and very objective factor to consider.

Many ways would be classified solely for being "part of a route" (no need show the same physical characteristics as the rest of the route), hence "(route)" in their suggested functional names. If a way is part of two routes of different class, the inferior route is considered to branch off from the superior route, so the way should be assigned the superior class. These rules also work well for routes across country borders.

Type	Functional name (could reduce mapper confusion)	Physical design	Minor destination¹	Remarks
highway=motorway	Freeway²	Controlled access²	Any
highway=trunk	National (route)	Any	If not assigned to an official system, metropolises (economics+population) and capital=* (politics)	Most countries assign trunks to some sort of official national highway system, regardless of physical characteristics (some are even unsealed)
highway=primary	Regional (route)		place=city	Most countries have their own definitions for these places, reflecting political status (understood as indicative of importance). In other cases, generic definitions based on population size are used. In urban areas, roads connecting suburbs are not assumed to reach a specific core, instead they are assumed to get anywhere in the suburb, sometimes just passing by, but in any case stretching multiple suburbs.
highway=secondary	Arterial (route)		place=town, place=suburb
highway=tertiary	Collector (route)		place=village, place=neighbourhood
highway=residential	(no change)		Residences when adjacent to the way	Usually only in urban areas. In rural areas, where most residences are in farms, use highway=unclassified. Considered less important than unclassified.
highway=unclassified	Local (way)		Anything else: place=locality, place=hamlet, place=isolated_dwelling, place=farm, landuse=industrial, landuse=commercial, etc.)

¹The opposite end destination should be of same or greater kind. For example: a tertiary connects two villages, or one village to another town, city, or metropolis. The destination is not an exact point, it is usually an area. Within that area, classification usually depends on continuity. For instance, secondaries within a suburb usually depend on routes connecting other suburbs rather than the suburb itself (it might be easier to begin with very distant suburbs), primaries within a city are those that connect the incoming primaries from the outside (the same would be valid for trunks; this usually leads to the formation of rings or arcs near the city core), etc.

²Etymologically misinterpreted in some places. "Free" refers to the fact that traffic never has to stop to wait for other traffic. All typical at-grade crossings of other highway types are replaced with overpasses and merging lanes.

Schematic diagram and general comments

Most cities develop approximately in a radial pattern, this is why a city formed by radiating ways and concentric circles and arcs is used as a basic, general model for classifying the streets of most cities. Cities planned as a grid will usually function in a radial pattern, with more activity at the core and radiating traffic to suburbs along preferential ways. As the city grows, important systems of rings and arcs naturally form as adjacent suburbs need to be connected to each other and incoming traffic needs to be routed around a congested city core. Systems of arcs that are not perfectly continuous sometimes provide the function of a continuous ring. Natural features often force distortions on the radial pattern. This includes rivers (before bridges are built over them), coastlines (straight, peninsular or bay-like) and terrain features (hills, mountains). Lack of urban planning may also produce such distortions, but it tends to fix itself over time for the higher levels (secondaries and above). Quite often on coastlines, the radial pattern will tend to elongate along the coast, resembling an ellipse instead of a circle.

Because motorways do not depend on their connections, they're not included in the illustration.

Please contact me if you're interested in modifying the OSM file that generated this illustration. There are no neighbourhoods 4 to 7 due to a typing mistake in "neighbourhood 8" (which should have been 4).

Note: this uses the old colour scheme. Yellow means tertiary (now white and wide), orange means secondary (now yellow), red means primary (now orange), green means trunk (now salmon), and blue would have meant motorway (now red).

Cities do not need to have actual suburbs and neighbourhoods, but they often form such structures. Suburb and neighbourhood nodes/centroids do not have to align with corresponding way crossings. For example, suburb 1 in the illustration is aligned, others are not. Actual suburb boundaries do not have to align with secondaries, but often do in some countries.

In organic suburbs and neighbourhoods, it is sometimes possible to match the highway mesh to a more or less grid-like subdivision of space. This often suggests which streets are tertiary or secondary when an official classification is missing or when surveying for rights of way and road attributes (width, surface, etc.) is not possible to the mapper.

If you can survey the area, then at least the following property must be valid: tertiaries always have right of way over residentials/unclassified, and secondaries always have right of way over tertiaries. When the area has poor signage, drivers' behaviour usually depends on physical characteristics such as road width and surface quality, with "ways with assumed priority" being better in both qualities.

The main route between two towns (for example, town 5 and town 6) is not always secondary, it is however always at least secondary. This suggests the following classification procedure for each road stretch:

• is it a trunk? yes if it connects two metropolises OR is part of the country's official trunk system agreed upon by the local community (e.g. a national highway system)
• otherwise, is it a primary? yes if it connects two cities
• otherwise, is it a secondary? yes if it connects two towns, or a town and a city
• otherwise, is it a tertiary? yes if it connects a village to something that is a village or higher (town or city)

etc.

Evaluating which ways are trunk based on metropolises may require knowledge of an entire continent, so it's best to have communities agree on a set of ways considered trunk and stick to that set. Primaries usually require a much smaller scope (countrywide or statewide), often simply looking at adjacent cities.

In densely built areas, the question changes from "connects" to "distributes to/from". For example, given a way in a neighbourhood:

• is it a primary? yes if it connects two city cores
• is it a secondary? yes if it distributes traffic directly (radially) from the city core into the suburb or between tangentially adjacent suburbs (this often forms a full ring or arcs approximating a ring)
• is it a tertiary? yes if it distributes traffic from nearby secondaries in and out of the neighbourhood

In "fringe" neighbourhoods, it makes little sense to declare a street a "tertiary" unless there is at least another parallel way for at least part of the way's extent.

Assigning many parallel roads the same very high class (like primaries, or even tertiaries) is usually an interpretation mistake. It is so because it doesn't help users choose their ways visually, nor make assertions about expected traffic (important at least for drivers, cyclists, and urbanists). That situation only makes sense if these ways belong to a greater, continuous system that is very high class. Since primaries connect cities, they would usually not appear parallel to each other; instead, they would be very long. Within the same city, specially in grid cities, many streets may connect multiple neighbourhoods, but still there is usually one or two that dominate; those should be secondaries, the others not. This creates a map where classification is sparse (easy to distinguish) at all zoom levels: most neighbourhoods are composed by many residentials and a few tertiaries, most cities are composed by many tertiaries, some secondaries and few primaries (towns would have some tertiaires, few secondaries and usually no primaries, unless they're in the route of two other cities), most states/countries are composed by many primaries and few very distinctive trunks/motorways.

This system does not require all classes to be present in any given area. For instance, the primary system could be entirely implemented by motorways, in which case no primaries are supposed to exist. This seems to be the case in Toronto.

Comparison to places with challenging classification

Note: the formation of primary rings and arcs is a consequence of connectivity between neighbouring cities through the city itself. Such arcs and rings are not a by-design feature, and indeed in some cities they may never form - there could be a single crossing of primaries at the city center. Primary rings and arcs may not form in case the city has a surrounding motorway ring which captures all through routes. Both scenarios are unusual but possible in principle. Because of that, a ring may be not entirely of the same class (for instance, primary). But a ring can be fully primary in certain cases, as an emerging consequence of routing between other cities. Additionally, as the city develops newer and better outer connections, inner primaries, including rings and arcs, may be downgraded to secondary.

City	Highway level disrupting regular radial pattern	Important influences on street pattern	OSM detail/community engagement	Significant changes upon current classification
Moscow	Secondary	River shape	Very high	The innermost and the outermost primary rings would be changed to secondary. The outer trunk ring would be changed into a primary if not part of the national highway system (but it probably is). See next section.
Paris	Secondary	River shape	Very high	None significant. This is remarkable considering that Paris is one of the most difficult cities to classify due to its irregular (though rectilinear) geometry. Like in Moscow, perhaps the outer trunk ring would be changed into a primary if not part of the national highway system (the wiki suggests that it has been elected a trunk by the local community). See next section.
London	Primary	River shape	Very high	Some OSM concepts seem to be applied differently in the UK. Probably some trunks would be downgraded to primaries, and some primaries (mostly short ones) downgraded to secondaries. This would actually make London resemble many other countries' big cities at first glance (such as Dublin, Paris and Berlin).
Berlin	Secondary	River shape	Very high	None significant.
Madrid	Secondary	River shape	High	Some of the current outer primary rings would be turned into secondaries.
New York City	Possibly none	Island shape	Medium	Probably many of the secondaries would be downgraded to tertiaries. Why are there so many highly concentrated secondaries there?
Los Angeles	Possibly none	River and coast shape	Low	Many primaries would become secondaries and many secondaries would become tertiaries. This is similar to the situation in NYC. Why? Is the local community not following functional classification, or is this a side effect from an import that's never been handled entirely? No other major online maps displays these roads so evidently.
São Paulo	Primary	River shape	Low	Most trunks would become primaries. Some trunks, along with primaries, both forming outer arcs, would become secondaries.
Singapore	Primary (see remarks)	Island shape, terran	Low	A majority of primaries would become secondaries, and some secondaries would become tertiaries. The current classification might be elevated due to the city's status as a "country". The motorway system suggests that the city has 3 important cores: the airport, downtown, and possibly the military areas on the west side. But even considering them as important as cities, most primaries would still become secondaries, particularly because their main connection is through the motorways. This, however, wouldn't be incorrect nor undesirable, and would still match map users' expectations trans-nationally.
Johannesburg	Secondary	Terrain	Low	None significant.
Cape Town	Primary (see remarks)	Peninsular shape, terrain	Low	None significant if the Table Mountain National park is considered as important as a city. Otherwise, many of the primaries on the peninsular side would be turned into secondaries, and some secondaries into tertiaries.

Ring roads

Modifying the idea of primaries as "connecting city cores" to "connecting city cores outside the built up area of the city, and inside connecting the incoming primaries" would usually classify most ring roads as primaries, and would also handle cases where no road is officially declared a "ring" road. In less developed areas, this could turn some unpaved/unsealed roads into primaries, which should be fine when this issue is solved.

The same idea extends to trunks, by replacing "cities" with "metropolises/capitals/entrypoints of national highways".

Criticism

There are several important advantages in this system compared to its alternatives:

• it produces a dense highway network even in undeveloped areas
• it takes into account both highway topology and road infrastructure
• it can work around infrastructure problems and incomplete/inadequate planning/construction work (including when parts of large roads have been built without being properly connected to the system)
• it shifts the focus of discussion to the question "what is the best route"; usually obvious, but there might be disagreement in less developed areas

Some have suggested that destinations such as airports, mines and tourist attractions (including some national parks) should influence the decision of road classification in certain situations. It can be argued that, when the number of visitors (tourists, workers, even somewhat permanent residents, etc.) over a period (such as the busiest day of week, at high season) is comparable to that of another settlement (typically 1k+ for villages/neighbourhoods, 10k+ for towns/suburbs, 100k+ for cities), that would justify increasing the road's class. Indeed, one could see permanent residents who commute (thus generating traffic) as similar to visitors of any kind (including then tourists and workers). Most of those situations would only cause roads to be promoted to tertiaries. One does not need to know surely and exactly how many regular visitors a destination has, just have an approximate notion of its order of magnitude (1, 10 or 100 thousand).

Without the idea of "number of visitors", a list of "important places" would have to be agreed on by the local community, and preferably written down in a common place. Foreign contributors might miss that workflow and end up making mistakes.

Possible necessary adaptations

• Illustrate how multiple primary rings can be formed by combination of routes between neighbouring cities. Demonstrate partial primary arcs as well, due to the same effect. Illustrate how "arcs" may form even on grid systems.
• Illustrate how classification should work on cities that form a single, continuous conurbation. In that case, there is probably a single primary connecting the two cities, and it is probably the connection between their cores.
• Illustrate extreme situations, such as:
  • A peninsular city with its core on the peninsula and two nearby cities, each with their own incoming primary route. Should the primary route inside the city lead all the way to the city core, or simply connect the entrances? What are the pros and cons? Is there a real example of this?
• Make illustration smaller to fit a mobile phone's screen while still being readable.

Adaptations whose importance is demonstrated on organic, developing or poorly planned areas

• It is assumed that structural characteristics, such as surface and width, will already affect the general idea of "best route" between places; as a consequence of political priorities and urban development vision, when similar options are available, the best route should always be the one with better infrastructure
• When deciding which routes are "the best/main" routes between places and which routes "the best/main distributors" of urban traffic locally, prefer the routes that are more usable by large vehicles (trucks, buses), which are usually large and with fewest sharp curves
  • This has implications for all forms of traffic (bikes will prefer to avoid those routes, cars can expect heavy traffic but reasonable flow, local dwellers - if any - can expect greater amount of noise)
• Routing applications that measure average speed can be used to argue in favor of one or another route, as long as the previous criteria regarding large vehicles takes precedence
  • It could be assumed that a place's assigned road class (primary for cities, secondary for towns, etc.) should always allow very large vehicles, such as the largest trucks; that means that, in a city, one would take such vehicles into account when classified primaries, and in a town, it would be accounted for on secondaries, and so on; there would be no need to classify secondaries in a city taking those large vehicles into account
• Very narrow ways can only be classified as tertiary (or higher) in very extreme and rare scenarios
• A "residential" street is often understood by urban planners as containing extra space between road and residence, so as to allow the circulation of pedestrians; many medieval towns and modern poorly planned cities have residential streets that do not fulfil that expectation; in those cases, it is useful to discriminate between wide residential ways and narrow residential ways, which are more similar to alleys
• Possibly related to central place theory's functional settlement hierarchy

Intention vs reality

• Especially significant in undeveloped/developing/unplanned areas
• Telling mappers to map what's on the ground and then defining rules for classification based on intention (trunks in Canada and Australia, footways vs path in most places) is contradictory
  • Intentions rarely change, while reality changes more often, especially in the form of disrepair
• Intentions might only be fully realised in reality after significant redevelopment of infrastructure (building/expanding and connecting ways)
• As such, though mapping intentions might be more useful in general, it may lead to confusing/counterintuitive situations, and as such may decrease the map's value
• It might make sense to forego the influence of intentions on classification when those intentions cannot be fulfilled in the foreseeable future, such as when concrete plans to implement the necessary redevelopment is still unheard of
  • Not the case of simple disrepair, but could be of significant/long-term disrepair

Urban ways in Germany (translated)

Original

Main roads

"Main road" cannot be strictly defined because it depends on the size of the place where the road is. "Residential" may suit small settlements [like towns], but national highways often pass through larger settlements [like cities]. Road class should transition into the local area without hopping, so the road at the entrance of a place normally takes the class of the road carrying inward. When a highway=unclassified is a town's main incoming road, it continues as "unclassified" inside the town.

In cities where primary roads have lost their previous significance, the strict rule of assigning primary to a "higher road" [a national highway] can be relinquished for roads that no longer belong to a higher road network. It may be useful to fill noticeable gaps with primary. Ring roads in major cities are sometimes classified as national roads, and then they are clearly mapped as "primary".

Thoroughfares

Thoroughfares are used primarily by through traffic (note: traffic that passes through the place, whose destination is elsewhere), but typically also have residents traffic. Their speed limit is usually 50 kmph, sometimes 60 kmph. A distinguishing feature of thoroughfares is better construction quality relative to residential ways, which can also have multiple lanes running in opposite directions. They usually have priority (right of way), better signage and improved intersections.

Multilane thoroughfares, which also serve cross-regional [nationwide] through traffic, are typically highway=secondary.

Thoroughfares with right of way, which lead into a neighbourhood or a residential area, are typically highway=tertiary.

Thoroughfares without right of way can be entered as highway=unclassified.

Residential ways

These roads in populated areas typically have:

• houses at least on one side
• small space for driving
• parking permission (sometimes only for residents)
• a separate pedestrian area (a sidewalk)
• negligible through traffic

In residential roads, priority to the right usually applies, as if entering first in a roundabout. Another criterion is that residential roads surround 30 kmph zones in privileged areas. Not strictly required when there are "Residents only" signs.

Greater roads with adjacent houses, with or without right of way, should be entered as thoroughfares (secondary or tertiary) or as main roads (primary).

Related discussions

• Urban highways in Vietnam: "classification is done by 'importance', not necessarily by physical attributs like the surface or width. A primary road might be unpaved in undeveloped countries"