After a bit of airline nightmare (do any US flights arrive on time anymore?) I made it to Limerick Ireland for OpenStreetMap’s (OSM) conference “State of the Map“. The talks have really highlighted how popular OSM has become. Roughly I break them into two buckets 1) the state of mapping a country 2) doing something cool with OSM data. The number of country presentations providing updates on OSM mapping progress is really impressive.

The usual suspects including Germany, Netherlands, France, our host Ireland and some surprising new places like Japan and Bolivia. The universal quality of each country presentation was how much had been mapped in the last year. The before and after pictures are quite dramatic. This was only reinforced by the massive posters on the wall of the conference venue showing six month incremental growth of mapped streets in the UK and Germany.

Equally impressive was the number of projects that are using OSM data for an innovative application. My personal favorite was OpenRouteService which allows users to geocode and route against OSM data. The project is being run at the University of Bonn and there are plans to open source the code, which would be great for providing open global geocoding services. Something we’ve struggled with finding. ITO also had a clever OSM implementation that allows you to query and filter through OSM data as separate map layers, both by feature type and edits (including temporal).

Wikitravel and Nestoria both had interesting examples of commercial services being built on top of OSM data. Wikitravel has a particularly cool approach where all the data they and the community has built up is free under Creative Commons, but they charge for on demand publishing of travel books you can take with you on a trip. Have to see if their geographic listings of travel amenities is something we can add to GeoCommons.

One of the most illuminating talks was “OpenStreetMap v the World” by Dair Grant, which provided a quantitative comparison of the accuracy of OSM vs. TeleAtals via Google Maps. Dair first did an analysis of Heath Scotland (roughly 10 km2) and found 89 errors in the TeleAtlas data. He then moved to Edinborough and did an analysis for a 10 km2 of OSM data and found 192 errors. Conclusion that none of the map providers are 100% accurate and OSM is not far off. A bit of detail on the OSM errors:

- 50% of the errors happened in 1km2
- 50% of the errors were missing roads (completeness is a common OSM challenge)
- 20% of the errors were missing names
- 15% of the errors were wrong names
- Only one error was from an incorrect junction

The big advantage of OSM when it comes to inaccurate data has been the ability to change the data easily. The big guys have caught on to this and just about all of them have developed technology to copy the concept:

- TomTom MapShare
- AND Map 2.0
- Google MapMaker
- TeleAtlas MapInsight

Dair pointed out that with all these services that there is no feedback loop to indicate your change was accepted and/or has been implemented. I believe this may not be true with Google MapMaker where the moderator provides feedback. Dair also provided a suggestion for OpenStreetBug to provide an easier mechanism to point out errors on the map. With all the discussion around crowdsourcing and accuracy this was a very enlightening talk. More to come from day two tomorrow.

Popularity: 19% [?]