GoKawiilOn 29-30 August, I attended the State of the Map (SotM) conference, in particular the scientific part. It’s been 15 years since the last time that I attended the SotM conference (last time 2011!), and it’s an opportunity to fill in a knowledge gap that I developed over this period. Unlike other conference reports that I’ve written, I am not summarising sessions, but capturing my impressions and aspects that I note through the renewed engagement with OpenStreetMap (OSM). The scientific part of the conference was particularly interesting for me, because it expresses the type of researchers that selected to present their work back to the community. Although the academic track is peer-reviewed and operates more like a scientific conference, the aim of the conference as a whole is more towards the community of OSM than the usual academic conference. OSM is used extensively in research – in 2025, OpenAlex suggests over 1250 papers, so I don’t expect that attending the session will be completely a review of what is going on. But the 20 or so papers do provide a notion of what is researched by the researchers who are closer to the community.
It is fortunate that I could attend SotM this year, considering that in June, I received the Test of Time award from the IEEE Pervasive Computing journal for the publication of the article on OpenStreetMap in 2008 (it is a top-cited paper in the journal), it is nice to get a sense of the papers that are citing it. In general, the academic/scientific track of the conference is doing well, with studies about OpenStreetMap and studies that use OSM data that filled the schedule for two days.
My first takeaway from SotM is that it was nice to see many familiar faces – there is a core group of people in OpenStreetMap that have been around now for about 20 years. For some, it is part of their career and what they do. Other people are doing it as a hobby in addition to their work. Either way, it is valuable to see how engagement can continue over such a long time. Secondly, unlike citizen science, there is much more presence of commercial actors – as sponsors of the conference, as presenters, and there was even an area for professional geospatial people who use OSM in France. This does provide resources, places of work for people who are in between enthusiasts and professionals (or professionalising their enthusiasm), and an engagement with the changing needs of the data.
Turning to the scientific track, from the start of scientific use of OSM, there were several characteristics that make it particularly attractive. It is an accessible, open, and hackable (in the sense that it is mutable and easy to understand) dataset. This makes OSM a site for experimentation in developing solutions to challenges such as routing, map generalisation, cartography, etc. However, it’s more messy data that needs to be examined, cleaned, and organised in order to use it for a specific investigation. And while the geometry might be complete, the attribute information continues to be hidden and variable. This messiness creates challenges for topology and routing – which makes it a persistent issue in the nature of the data produced. It is valuable to note how routing remains an area of experimentation and challenges.
But there are plenty of things to map: for example, attribute completeness for dams in rural spaces is very low – below 1%. There is also interest in indoor mapping and completing details of public buildings. Of course, the world continues to change, and you need to understand where and how changes are happening. The emergence of multiple open geospatial data sources is making it possible to keep the OSM approach to the use of the data. Satellite imagery continues to play an important part as a source of information. Another area for improvement in mapping can be the indication of building entrances instead of centroids – which is very relevant for navigation applications.
An example of the ongoing research on data quality is the exploration of completeness – using extrinsic data comparison, intrinsic attribute analysis, or statistical estimation. The methodology that was developed combines intrinsic attributes and statistical methods to evaluate completeness – assuming that there are features that will be captured first (say roads) and things that will be saturated at the end (say addresses) it is possible to check over time to see how features are being added until the map stabilises. The analysis of completeness in this way is relevant for a specific class of feature (or attribute) – so the question can be: is this a complete set of buildings? etc.
In terms of the application areas that OSM data is being used on, there were examples from public health studies. OSM is considered relevant enough to explore if it can provide information on rural spaces (which wasn’t the case in the past). A similar example is applications in monitoring mining activities across the world. There are also new problems that need addressing – such as mapping the electricity grid (my very first large project in GIS was on digitising the mapping of the Israel Electric Company in 1991). Interestingly, the quality assurance of OSM is seen as valuable – with tools such as osmose. For the grid, consistency, completeness, and up-to-dateness are core parameters (in mapmygrid.org/quality).
What is also interesting is that because of the good level of completeness – especially in large urban areas- there are increasing large scale studies that use OSM as a basis for analysis. This can be included in the issue of data quality – a persistence issue.
The role of OSM as a humanitarian source of mapping in places where information is missing continues. On the practical side, it is an effective and efficient way to produce maps, and there are even evaluations of the low costs that such mapping involves.
I was somewhat surprised to see that most of the examples that were shown didn’t use other open data projects and merged the data for evaluation and analysis. One of the only examples was the use of the Colouring Cities project that is running from the Touring Institute. Since my early days in geospatial research, I am baffled, and continue to be, about different analyses that are in the form “we try to solve problem X only with data from source Y”. For example, research on road lanes, or the characteristics of an urban park that only uses OSM without using other sources. I think that one of the major reasons that it continues to be the case, almost 30 years later, is the learning costs of getting familiar with a data source and knowing how to use it. PhDs, postdoc fellowships, or research projects are always limited in time, and it probably feels like the effort of learning all the ways in which you should use a dataset is time-consuming and complex. There is a lot of trial and error, so you stick to one source. Yet, maybe the thing that people should do is to reduce the geographical scope of their question while trying to explore multiple sources of information. There is so much open data of high quality out there – from Wikipedia, OpenStreetMap, Satellite data, Citizen science data, etc. Creative approaches to merging and using different data sources might be a more effective way to answer the question…
... continue reading