This post follows on from some of the questions I had back in January about using GPS traces to create road maps.
The average smartphone GPS is accurate to about 5 metres, and that accuracy degrades near steep mountains, or in built-up areas with tall buildings. You can see this when trying to trace out the contours of a relatively small area, like a village playground. GPS on a smartphone stops being useful for measuring any area smaller than about 20 metres squared.
Meanwhile I don’t know how wide British roads are on average, but various road design guidelines seem to settle on 3 metres as the recommended minimum lane width.
In this case we have a problem, the margin of error given by our measuring instruments is wider than the thing we’re trying to measure.
So, what does a volunteer cartography project do? OpenStreetMap has collected around 7.5 million GPS points (at the time of writing), and even if most of those are a bit off, the overall volume of repeated measurements will eventually converge on a ‘generally correct’ picture of the road network. This is the same idea behind the volunteer-edited map too. There is no single definitive map, only a consensus built out of collaboration.
The opposite approach would be to get a team of professionals to map out roads using expensive and super-accurate equipment. This is the Encyclopedia Britannica vs. Wikipedia scenario - verification by authoritative experts over the collective wisdom of the crowd.
Which approach results in the most accurate measurement of the road network?
The Earth moves
If you want to go really far into this, you have to acknowledge that the Earth’s surface is not static. Over enough ground and with enough time, the Earth’s tectonic plates will drift out of alignment with the fixed points on the map.
Australia is one of the fastest-moving areas on the globe, it moves by about 6.5 cm per year. That means any measurements made in Australia ten years ago now have a margin of error of up to 65 cm. In ten years time that margin of error will grow to 1.3 metres, relative to the rest of the world.
Earthquakes also cause disruptive tectonic motion, which squeezes and stretches the earth’s surface. They’re not easily predictable, and they move about large chunks of land in an irregular way.
There’s nothing that amateur geographers can really do about the earth moving, but what was once considered a joke is a slowly developing problem for anyone who cares a lot about keeping a global map accurate down to a metre.
Is a road a line?
There’s another problem with measuring road widths. Roads are lines on maps, but on the ground they cover an area. You can see this problem when lines and areas overlap.
Here’s an example I’m very familiar with, Magdalen Bridge in Oxford.
There’s a gap immediately noticeable between the road (mapped as a line) and the pavement (mapped as a polygon area). Even worse, the pavement also has a footpath running through it, what a mess.
In principle the only thing that matters is that the underlying data is accurate, rendering glitches are supposed to be irrelevant. And yet, this data is technically being rendered correctly, the problem is in the way the bridge is represented in the OSM database.
Here’s another awkward area, Broad Street in Oxford.
The road in this case stops at the pedestrian area at the end of the Broad Street bottleneck.
Here’s how it looks in the Google
Maps Earth satellite view.
And from ground level.
You can see that in reality the road is not a single lane, it’s the continuation of a wide open road area. From the perspective of a pedestrian, this area could almost be a public square. It’s so wide that over winter it gets repurposed as the site for a Christmas Market. On the other hand, from the perspective of a car this is not a free-for-all road surface, there are clear rules about circulation.
Magdelen Bridge and Broad Street don’t fit the descriptive model of the map. How do we best represent spaces like these?