Some maps, tables and charts on the distribution and length of sealed (aka paved) and unsealed major roads (from motorway to tertiary) in all states of Australia.
Early this year, after editing roads near my home, I looked further afield and realised that, if some gaps on OpenStreetMap (OSM) were filled, virtually all major roads in Australia could soon have a surface tag to show whether they were paved or unpaved.
Like thousands of mappers before me, I pottered away and filled some gaps. Gradually, only a few dozen roads remained without surface tags in most states; usually where I couldn’t determine the surface type (e.g. the road was under dense trees or wasn’t visible on available imagery). Having said that, two huge gaps defeated me: in Perth and Melbourne, but both are in urban areas and nearly all of the untagged roads in those areas will be paved.
At the “end” of the process, when I looked back over the areas I’d worked on earlier, more new roads had been added to the map, and many of those don’t have surface tags either. OSM never stands still. Nevertheless, we’ve reached the stage whereby it’s possible to say that, with localised exceptions, nearly all major roads across Australia now have a surface tag on OSM. That’s an amazing accomplishment by the mapping community, and a great resource for anyone who drives a truck or tows a caravan around the country and who wants to stick to the tarmac and avoid the dirt.
So what do all those tags show? To find out, I collated the highway surface data from OpenStreetMap to explore the patterns of sealed and unsealed major roads across Australia. The results are shown below. I hope you find something of interest.
Throughout this post I use the term major road to refer to the five uppermost highway classes in OpenStreetMap: highway = (1) motorway, (2) trunk, (3) primary, (4) secondary and (5) tertiary, as described on the international wiki and the Australian tagging guidelines. Virtually all roads in these five classes now have a surface tag (usually asphalt, paved, or unpaved). All other highway classes (unclassified, residential, service, link roads, track, etc.) are excluded, as many ways do not have surface tags.
The term sealed (instead of paved) is used as an umbrella term for all accepted surface tags that refer to sealed/paved roads, including surface=paved, asphalt, concrete, etc. This avoids ambiguity over whether “paved roads” refers only to highway=paved or to highway=paved+asphalt+concrete, etc. Similarly, unsealed includes all accepted tags for unsealed/unpaved roads, including surface=unpaved, gravel, dirt, sand, etc.
For international readers, Australian states are abbreviated as follows: Australian Capital Territory (ACT), New South Wales (NSW), Northern Territory (NT), Queensland (Qld), South Australia (SA), Tasmania (Tas), Victoria (Vic) and Western Australia (WA).
Skip ahead if you just want the tables and charts.
In all analyses, I have accepted all mapped tags at face value. This includes hierarchy tags (motorway, trunk, etc.) and surface tags (paved, asphalt, unpaved, etc.). Many mappers are editing both groups of tags and the patterns displayed here will change as tags continue to be refined. Treat the results as a summary of the “state of play” rather than a fait accompli.
Overpass Turbo was used to calculate road lengths and create all maps. All queries were made on 15 June 2021. I collated data state-by-state and added the state figures to generate national data. First, I used Overpass Turbo to calculate the total length of all roads (sealed, unsealed and untagged) for each hierarchy level in each state (A). I then repeated the process to calculate the total length of unsealed roads for each hierarchy level in each state (B). I estimated the total length of sealed roads in each state by subtracting B from A. Thus, the total length of sealed roads = the total length of all roads minus the total length of unsealed roads.
I used this indirect approach instead of using Overpass to directly calculate the length of sealed roads for a simple reason. The two major groups of roads with no surface tags are in urban Perth and Melbourne. I have assumed that all of these roads are sealed, and this approach lumps untagged roads with sealed roads. I’ve also assumed that the majority of major roads without surface tags that others added to OSM after I “finished” ( lol ) my edits in each State were also sealed. With the exception of a few long mine access roads in NT, SA and WA, this assumption is likely to be largely true. Regardless, the impacts of incorrectly grouping some untagged roads with sealed roads will be trivial compared to the big patterns described below.
Many roads with dual carriageways are mapped using parallel ways, one running forwards and one backwards. If one calculates the total length of these roads then the result will be double the expected outcome. To illustrate, imagine a motorway runs 100 km from town A to town B. If the road was mapped as a single way (as on most roads), the straight-line length of the road would be 100 km. However, for dual-carriage way roads, the result will be 200 km, as the map includes the distance from town A to B, and then back again from B to A. I haven’t compensated for this effect, but draw attention to it in a couple of places. This will mostly affect the estimated lengths of sealed motorways and some trunk, primary and secondary roads. It will have no impact on the distance of unsealed roads.
First, the big picture. In total, according to OpenStreetMap data, there are nearly 324,000 km of major roads across Australia.
Source: OpenStreetMap, 15 June 2021.
Two states contain half of these roads: NSW (nearly 85,000 km, 26% of the total length of all major roads) and Qld (79,000 km, 24%). At the other extreme, the ACT, NT, SA and Tas each contain less than 10% of the total, with the ACT accounting for 0.5% of the total length of major roads in Australia (see Table 1 & Map 4).
Not surprisingly, the total length of all motorways is shorter than that of all trunk roads, which in turn is shorter than that of all primary roads, and so on (see Chart 1). Thus, motorways account for just 2% of the total length of major roads (hence their “real” length is probably about 1% of all roads when dual lanes are accounted for). By contrast, tertiary roads account for 43% of the total length of major roads.
This broad trend (i.e. trunk roads are shorter than primary roads, which are shorter than secondary roads, etc.) also occurs in most states, with two minor exceptions (see Table 1 above). In WA, the total length of primary roads (just under 7,000 km) is less, not more, than that of trunk roads (nearly 12,000 km). In Tasmania, the total length of tertiary roads (nearly 1,300 km) is much less than that of secondary roads (about 4,500 km). While atypical at the state scale, the Tasmanian pattern is not unusual in many regions on the mainland.
In total, 70% of the total distance of major roads across Australia is sealed (about 227,000 km) and only 30% is unsealed (about 97,000 km). It would be interesting to compare these percentages if urban roads, nearly all of which are sealed, were excluded and rural roads only were analysed.
The proportion of major roads that is unsealed varies greatly across states (see Map 5 & Table 2). In the NT, over half of the total length of all major roads is unsealed (55%). At the other extreme, only 5% of the total length of major roads in Victoria is unsealed and just 3% in the ACT. In fact, only 53 km of tertiary roads in the ACT is unsealed.
Source: OpenStreetMap, June 2021.
From the figures shown in Table 2 above, one can divide the eight states into four groups, to give an even broader view of the patterns of sealed and unsealed roads across road hierarchy classes.
|1||NT||Many trunk roads are unsealed (17% of total length).|
All road classes have a higher proportion of unsealed roads than in other states.
Nearly all tertiary roads are unsealed (91% of total length is unsealed).
|2||WA & Qld||Virtually all trunk roads are sealed.|
A small but sizeable proportion of primary roads is unsealed (27 & 13% respectively).
In both states, just over one-third of the total length of major roads is unsealed (36%).
|3||SA, NSW & Tas||Virtually all trunk and primary roads are sealed.|
Roughly half of the total length of tertiary roads is unsealed (46-56%).
|4||Vic & ACT||All trunk, primary and virtually all secondary roads are sealed.|
At least 90% of the total length of tertiary roads is also sealed.
While not analysed here, most unsealed rural roads must be tagged as unclassified or tracks.
Broadly, as one moves down the road hierarchy, the proportion of roads that is unsealed increases (see Chart 2). All motorways are sealed (by definition). Only 2% (nearly 700 km) of the total length of trunk roads is unsealed; most of this is in the NT on the Central Arnhem Highway (583 km unsealed: see Map 6). The remainder is in WA on the Goldfields Hwy between Meekatharra and Wiluna. At the other end of the road hierarchy, nearly half of the total length of tertiary roads across Australia is unsealed (48%).
If one lumps motorways and trunk roads together (as in Chart 2 above), the total length of motorway+trunk roads (just over 48,000 km) is almost the same as the total distance of all primary roads (nearly 47,000 km). However, a slightly greater proportion of primary roads are unsealed (as shown by the light blue bar in Chart 2). The total length of secondary roads (nearly 90,000 km) is almost double that of primary roads, and the extra roads include an even mix of sealed and unsealed roads. Not surprisingly, the total length of tertiary roads is longer again (over 139,000 km), but this increase is mostly due to an influx of unsealed roads. Curiously, the total distance of sealed roads is similar for both secondary and tertiary roads (about 66,000 and 72,000 km respectively).
Regional and local patterns
So far, I’ve focused on state and national patterns, but the detailed data in OSM is usually most useful at local levels. Local and regional maps of road surfaces help routers and road users to design routes that avoid unsealed roads (“stick to the tarmac”) or, alternatively, to target unsealed roads. Truck drivers prefer bitumen and gravel bike riders like dirt.
The usefulness of surface tags on major roads varies depending on where you live. Virtually all major roads in and around large cities are sealed, so surface tags may be viewed as superfluous. In more remote areas, some primary, secondary and tertiary roads are sealed and others are unsealed, so surface tags can help road users select their preferred routes. Map 7 (below) shows an area in central Queensland with a wide mix of sealed and unsealed primary, secondary and tertiary roads. Accurate surface tags are most useful in regions like this.
In areas where most tertiary roads are sealed (as in Victoria), surface tags will be most useful if unclassified roads (the hierarchy level below tertiary roads) are also tagged comprehensively. I know of two large regions with a high road density where all unclassified roads have surface tags: north-west Victoria (where most unclassified roads are unsealed) and the Riverina region in southern NSW (where unclassified roads contain a mixture of sealed and unsealed surfaces). Comprehensive tagging in the Riverina makes it possible to create routes that almost completely avoid sealed roads (see Map 8, for example).
Stable or ephemeral?
New roads and tags continue to be added to OpenStreetMap, which raises the question: how stable or ephemeral are the patterns displayed here?
I suspect that most patterns will remain relatively stable. At a fundamental level, the enormous number of roads that are already mapped in OpenStreetMap will simply overwhelm the impacts of future edits. Ongoing editing will improve the accuracy and utility of OSM at local and regional scales but will operate within the superb framework that was established over the past decade.
The patterns shown for motorway, trunk and primary roads are likely to be the most stable. High-level roads are well mapped and tagged and nearly all are already sealed, so there is little opportunity for substantive changes to the patterns that already exist. Secondary roads are also well mapped. Currently, 26% of their total length is tagged as unsealed. Some of this is undoubtedly due to out-of-date data in OpenStreetMap, and many sections will be upgraded from unsealed to sealed as edits (and on-ground works) continue. The patterns shown for tertiary roads may be the least stable, largely because the allocation of roads to tertiary, rather than to secondary or unclassified classes, varies greatly between and within regions, and road hierarchy tags may be subject to considerable revision in the future.
Nevertheless, while road hierarchy and surface tags will continue to be refined on individual roads, the sheer size of the Australian road network, and the immensity of existing mapping, suggests that the broad patterns displayed here are likely to persist. Perhaps the greatest opportunity for increased utility will come from ongoing mapping of road surfaces on unclassified roads, especially in regions where unclassified roads contain a mix of sealed and unsealed surfaces.
Climate, landscapes, human populations and landuse vary enormously across Australia. The road network reflects all of these factors, and patterns of sealed and unsealed roads are important features for road users. Beyond the obvious utility of OpenStreetMap for map users, it’s just plain fun to uncover patterns that are “hidden” in tags and not shown on many map renders. I hope you found something of interest in this babble of charts, tables and maps.
Finally, I can’t help but wonder sometimes; just what did the first, small group of mappers really think they could create, when they mapped, named and added tags to the first bunch of roads somewhere in Australia? Where were those first roads? As a Johnny-come-lately, I’m privileged to stand on the shoulders of the OpenStreetMap community. Everything here is a testament to the community’s enormous work. To everyone who has contributed, you’ve left an extraordinary legacy. Many thanks.