Digital Finds

joseph dot reeves at thehumanjourney dot net
@iknowjoseph

Route finding across a continent: Testing changes to OSM data

Mar 15, 2010 by Joseph Reeves

There's currently an impressive OpenStreetMap tidy-up effort quietly happening across the globe, you can read about it here. Of course, duplicate nodes aren't the only sort of messy data possible to see on OpenStreetMap, but it's a good place to start if you want to help improve the quality of data within the database. Personally, I'm most interested in mapping Africa, so I'm always keen to see how many red crosses are on this view. As it happens, we can look at how this tidy-up, and others, have impacted the quality of OSM data by trying to plot a route with CloudMade across the continent, but first some more introduction.

Mapping in Africa was given a fantastic boost relatively recently by Robert Soden at Development Seed when he was responsible for importing over a hundred thousand miles of Africover road data into OSM. Obviously this is absolutely fantastic, but I was finding that when clearing up dupe nodes, and looking at aerial imagery, that the Africover data was particularly liable to errors and broken segments. This isn't really a problem; OSM is a wiki and we're all able to tidy up this data - think of these roads as stubs and we've got an amazing starting point from which to build a free map of the continent. Tidying Africover data, and ways sourced from elsewhere, I began to think that it must be possible to measure improvements made to the map. The thinking is simple: You plot a route from A to B, the produced route isn't perfect but you note the calculated distance; you then improve the data and the route between A and B becomes shorter as the software you use doesn't have to route around errors and can plot a more direct route.

Such thinking isn't exactly new; people have long been using routing applications to test the quality of OSM data. You can do it now - try entering your home and your place of work into a routing application and seeing if it comes up with the way that you actually travel. If not, you might go home a funny way, or OSM might have some incorrect data about your part of the world. Regardless, if you record the same route at intervals across a period of time and you could start to judge how the OSM data in a region is changing.

With that thought in mind I was very pleased to revisit an old blog post at 27 Months that spoke of routing from Cape Town to Ethiopia; most importantly it contained a screenshot (reproduced below as I didn't want to leech anyone's bandwidth) and a link defining the start and end point. All we have to do is click the link and compare the results:

Before:

After:

In August 2009 the routed distance is recorded as 7298 km, taking in the south west and central countries of Botswana and Namibia; by March 2010, the route had shrunk to 6666 km and travelled south through Zimbabwe. This seems to show a great improvement in the quality of road data for south east Africa; as the recorded roads get more complete and contain less errors, shorter routes can be plotted along them. CloudMade allow you to download the route as a GPX file; a potentially very useful record of the state of OSM data. Also, note that in the first image you can read Zaire, something Wikipedia tells me hasn't existed since 1997; a further indicator that data improves over time.

The above image shows us that we should perhaps be careful; shown on the map is a route from Morocco to Algeria, although the guide books tell us that this border crossing is closed. At least the data can be easily changed, however, and routing services used to check the accuracy.

Good news all round then I think: Routing services can be used to provide an analysis of OSM data over time (as long as you can find historic examples), it seems to be improving and, if errors remain, routing tools can be used to improve and verify road information across an entire continent. Whilst I've only demonstrated the concept here, if we compare regularly generated GPX files from CloudMade at some point in the future, we will likely be able to say a great deal about road coverage in the region; continuing to look at this in the future may reveal insights into the ongoing efforts to map Africa or elsewhere.



Comments:

Great comparison, Joseph. Clearly, the quality of OSM data for E. Africa has improved substantially over a relatively short period. Here's a larger version of the Cape Town to Ethiopia map from the original post. Nice catch with Zaire, by the way. ;)

Posted by Bill Zimmerman on March 16, 2010 at 09:46 AM GMT #

Routing OSM from Most Southern Tip to Most Northern Tip of Africa:
http://bit.ly/c7XkOp

The DRC section needs some work, but the route follows the preferred route along the Trans-African Highway:
http://en.wikipedia.org/wiki/Trans-African_Highway_network

Posted by Firefishy on March 16, 2010 at 11:39 AM GMT #

Post a Comment:
Comments are closed for this entry.