Maps are increasingly becoming more and more popular on web pages, either embedded in existing pages or as dedicated pages. Additionally new tools featuring advanced analytical functionality are also on the up and coming. This increase in web-maps has spawned a great need for geographical data, from basic information, such as administrative boundaries to more detailed information such as hiking trails and similar, in addition there is the information space which holds geo-referenced information, such as images, people and similar.

In a response to this need several large map sites offer the ability to add/change (and even delete) the geographical information seen in the map such as Open Street Map (OSM) and Google Map Maker to mention a few.  Google Map Maker base itself on commercial made information, either freely available or purchased in addition to the contributions from the users, Open Street Map on the other hand base itself solely on user contributions – both approaches has to deal with the accuracy of the data. For the commercial data the accuracy is believed (or guaranteed) to be of some level, often acceptable for the regular user, for the user contributed there are no accuracy provided, however it is believed that it is correct. This problem of accuracy is starting to emerge quite rapidly as the screenshots below indicates. Which map holds the correct information? Well probably it is the one with the most details and with the least jagged lines – and yes, the N50 map is the most semantically correct – not surprising since it is the Norwegian Mapping Authority which is responsible for it. However, how should one know this by just looking at, for instance, the Open Street Map? Or even worse the Yahoo Map?

Comparison between different web maps. From the left; Yahoo Maps, Open Street Map, Gulesider.no, N50 (Norwegian Mapping Authority)

Here is where the legacy of maps come in to play. Maps have an enormous trust among the users – people generally trust their lives with maps and strongly believe that what the map depict is the truth – however that may not be true, at least not always.  The example shown here is quite harmless, but what if it was a reef in the ocean that was slightly jagged in the map, but in real life stretched further? Such accuracy is not communicated with todays maps, at least not as explicitly as it should. Although I believe for the larger part the accuracy is known, probably not in centimetres or meters, but at least in the sense correct/medium/false or similar. The advent of participatory map making communities/tools is also posing in favour of this, as the community as such could rank the semantic validity of the information (this may already exist in for instance OSM?).

So, I suggest that more effort is put into the explicit communication of the accuracy of information communicated in maps.

Some solutions that could support this is either to avoid inaccuracy in the information. This is done at a large scale already by aggregating data (arithmetic average/weighted average etc), filtering from sources (user contributions vs. commercial data), or rating from users. However it is inevitable that inaccuracy occurs, so allowing it (or embracing it) is better than avoiding it. If one is to allow for maps with possibly large inaccuracies, then one must have a way of handling it and presenting it to the user in such way that the user benefits from it and not the opposite. I believe one such method lies in the presentation of data – if it is utmost clear that the data may be inaccurate, then the user can freely decide what he will do with it (trust it or just “keep it in mind” etc). Techniques for this may be;

Wiggly shapes
Visualize inaccurate data with wiggly shapes to indicate uncertainty. For instance the jagged lines in the Yahoo Map could be made “wiggly” indicating that “we know there is a shore here, it goes something like this, but we are not sure“.

Buffer zones
Adding “buffer zones” to the shapes to indicate the uncertainty of the data. This could be very well suited for commercial data where the inaccuracy often is given/calculated on the basis of the measurements performed. However, buffer zones could potentially be used to indicate more loose uncertainty also – but probably suffers from the legacy trust that maps in general “suffers” from, i.e. that the zone is perceived as correct knowledge about the inaccuracy.

Colour coding

One of the more intuitive solutions is to colour code inaccurate data with special colours, for instance red for high inaccuracy and green for low inaccuracy. However the map would probably suffer from a cognitive information load that would make the map more or less useless.

Ghost maps

Several overlapping data sets is probably (or will be) the case of many map applications in the future, such as Open Street Map and commercial maps with the INSPIRE directive implemented. The data sets may vary greatly and instead of choosing one, or aggregating some of them, one solution may be to lay them on top of each other with a varying degree of transparency or even hollowness. This would communicate all information at hand to the user, whether right or wrong. Additionally the accuracy could be implicitly decided by the user as he would perceive the data and find similarities (or dissimilarities) among the data by using visual queues alone.

Text labeling

The most obvious communication of accuracy is to convey it using text. Either in the map or in popup information boxes or similar. This is fairly dull as I see it. The method requires to much effort from the user and fails at providing an overview over a large number of entities and their accuracy. As such, this method will not provide a valid solution.

Well, that was some of my ideas:) I’d love to hear your thoughts on this subject! Is the problem really a problem or are maps working good with their legacy trust and legacy cartography? Has something like this been dealt with before? Are there any related research on this?

• XBox Natal – challenge to develop the first GIS app.
• Extending the magnifier-metaphor
• Google Buzz – managing redundancy