Following the last post, which focused on an assertion about crowdsourced geographic information and citizen science I continue with another observation. As was noted in the previous post, these can be treated as ‘laws’ as they seem to emerge as common patterns from multiple projects in different areas of activity – from citizen science to crowdsourced geographic information. The first assertion was about the relationship between the number of volunteers who can participate in an activity and the amount of time and effort that they are expect to contribute.
This time, I look at one aspect of data quality, which is about consistency and coverage. Here the following assertion applies:
‘All information sources are heterogeneous, but some are more honest about it than others’
What I mean by that is the on-going argument about authoritative and crowdsourced information sources (Flanagin and Metzger 2008 frequently come up in this context), which was also at the root of the Wikipedia vs. Britannica debate, and the mistrust in citizen science observations and the constant questioning if they can do ‘real research’.
There are many aspects for these concerns, so the assertion deals with the aspects of comprehensiveness and consistency which are used as a reason to dismiss crowdsourced information when comparing them to authoritative data. However, at a closer look we can see that all these information sources are fundamentally heterogeneous. Despite of all the effort to define precisely standards for data collection in authoritative data, heterogeneity creeps in because of budget and time limitations, decisions about what is worthy to collect and how, and the clash between reality and the specifications. Here are two examples:
Take one of the Ordnance Survey Open Data sources – the map present themselves as consistent and covering the whole country in an orderly way. However, dig in to the details for the mapping, and you discover that the Ordnance Survey uses different standards for mapping urban, rural and remote areas. Yet, the derived products that are generalised and manipulated in various ways, such as Meridian or Vector Map District, do not provide a clear indication which parts originated from which scale – so the heterogeneity of the source disappeared in the final product.
The census is also heterogeneous, and it is a good case of specifications vs. reality. Not everyone fill in the forms and even with the best effort of enumerators it is impossible to collect all the data, and therefore statistical analysis and manipulation of the results are required to produce a well reasoned assessment of the population. This is expected, even though it is not always understood.
Therefore, even the best information sources that we accept as authoritative are heterogeneous, but as I’ve stated, they just not completely honest about it. The ONS doesn’t release the full original set of data before all the manipulations, nor completely disclose all the assumptions that went into reaching the final value. The Ordnance Survey doesn’t tag every line with metadata about the date of collection and scale.
Somewhat counter-intuitively, exactly because crowdsourced information is expected to be inconsistent, we approach it as such and ask questions about its fitness for use. So in that way it is more honest about the inherent heterogeneity.
Importantly, the assertion should not be taken to be dismissive of authoritative sources, or ignoring that the heterogeneity within crowdsources information sources is likely to be much higher than in authoritative ones. Of course all the investment in making things consistent and the effort to get universal coverage is indeed worth it, and it will be foolish and counterproductive to consider that such sources of information can be replaced as is suggest for the census or that it’s not worth investing in the Ordnance Survey to update the authoritative data sets.
Moreover, when commercial interests meet crowdsourced geographic information or citizen science, the ‘honesty’ disappear. For example, even though we know that Google Map Maker is now used in many part
s of the world (see the figure), even in cases when access to vector data is provided by Google, you cannot find out about who contribute, when and where. It is also presented as an authoritative source of information.
Despite the risk of misinterpretation, the assertion can be useful as a reminder that the differences between authoritative and crowdsourced information are not as big as it may seem.
5 December, 2012
Recently, I attended a meeting with people from a community that is concerned with vibration and noise caused by a railway near their homes. We have discussed the potential of using citizen science to measure the vibrations that pass the sensory threshold and that people classify as unpleasant, together with other perceptions and feeling about these incidents. This can form the evidence to a discussion with the responsible authorities to see what can be done.
As a citizen science activity, this is not dissimilar from the work carried out around Heathrow to measure the level of noise nuisance or air pollution monitoring that ExCiteS and Mapping for Change carried out in other communities.
In the meetings, the participants felt that they need to emphasise that they are not against the use of the railway or the development of new railway links. Like other groups that I have net in the past, they felt that it is important to emphasise that their concern is not only about their locality – in other words, this is not a case of ‘Not In My Back Yard’ (NIMBY) which is the most common dismissal of local concerns. The concern over NIMBY and citizen science is obvious one, and frequently come up in questions about the value and validity of data collected through this type of citizen science.
During my masters studies, I was introduced to Maarten Wolsink (1994) analysis of NIMBY as a compulsory reading in one of the courses. It is one of the papers that I keep referring to from time to time, especially when complaints about participatory work and NIMBY come up.
Inherently, what Wolsink is demonstrating is that the conceptualisation of the people who are involved in the process as selfish and focusing on only their own area is wrong. Through the engagement with environmental and community concerns, people will explore issues at wider scales and many time will argue for ‘Not in Anyone’s Back Yard’ or for a balance between the needs of infrastructure development and their own quality of life. Studies on environmental justice also demonstrated that what the people who are involved in such activities ask for are not narrow, but many times mix aspects of need for recognition, expectations of respect, arguments of justice, and participation in decision-making (Schlosberg 2007).
In other words, the citizen science and systematic data collection are a way for the community to bring to the table evidence that can enhance arguments beyond NIMBY, and while it might be part of the story it is not the whole story.
For me, these interpretations are part of the reason that such ‘activism’-based citizen science should receive the same attention and respect as any other data collection, most notably by the authorities.
Wolsink, M. (1994) Entanglement of Interests and Motives: Assumptions Behind the NIMBY-Theory on Facility Siting, Urban Studies, 31(6), pp. 851-866.
Scholsberg, D. (2007) Defining Environmental Justice: Theories, Movements, and Nature. Oxford University Press, 2007
In March 2008, I started comparing OpenStreetMap in England to the Ordnance Survey Meridian 2, as a way to evaluate the completeness of OpenStreetMap coverage. The rational behind the comparison is that Meridian 2 represents a generalised geographic dataset that is widely use in national scale spatial analysis. At the time that the study started, it was not clear that OpenStreetMap volunteers can create highly detailed maps as can be seen on the ‘Best of OpenStreetMap‘ site. Yet even today, Meridian 2 provides a minimum threshold for OpenStreetMap when the question of completeness is asked.
So far, I have carried out 6 evaluations, comparing the two datasets in March 2008, March 2009, October 2009, March 2010, September 2010 and March 2011. While the work on the statistical analysis and verification of the results continues, Oliver O’Brien helped me in taking the results of the analysis for Britain and turn them into an interactive online map that can help in exploring the progression of the coverage over the various time period.
Notice that the visualisation shows the total length of all road objects in OpenStreetMap, so does not discriminate between roads, footpaths and other types of objects. This is the most basic level of completeness evaluation and it is fairly coarse.
The application will allow you to browse the results and to zoom to a specific location, and as Oliver integrated the Ordnance Survey Street View layer, it will allow you to see what information is missing from OpenStreetMap.
Finally, note that for the periods before September 2010, the coverage is for England only.
Some details on the development of the map are available on Oliver’s blog.
How Many Volunteers Does It Take To Map An Area Well? The validity of Linus’ law to Volunteered Geographic Information
10 January, 2011
The paper “How Many Volunteers Does It Take To Map An Area Well? The validity of Linus’ law to “ has appeared in The Cartographic Journal. The proper citation for the paper is:
Haklay, M and Basiouka, S and Antoniou, V and Ather, A (2010) How Many Volunteers Does It Take To Map An Area Well? The validity of Linus’ law to Volunteered Geographic Information. The Cartographic Journal , 47 (4) , 315 – 322.
The abstract of the paper is as follows:
In the area of volunteered geographical information (VGI), the issue of spatial data quality is a clear challenge. The data that are contributed to VGI projects do not comply with standard spatial data quality assurance procedures, and the contributors operate without central coordination and strict data collection frameworks. However, similar to the area of open source software development, it is suggested that the data hold an intrinsic quality assurance measure through the analysis of the number of contributors who have worked on a given spatial unit. The assumption that as the number of contributors increases so does the quality is known as `Linus’ Law’ within the open source community. This paper describes three studies that were carried out to evaluate this hypothesis for VGI using the OpenStreetMap dataset, showing that this rule indeed applies in the case of positional accuracy.
To access the paper on the journal’s website, you can follow the link: 10.1179/000870410X12911304958827. However, if you don’t hold a subscription to the journal, a postprint of the paper is available at the UCL Discovery repository. If you would like to get hold of the printed version, email me.
29 November, 2010
The website GPS Business News published an interview with me in which I covered several aspects of OpenStreetMap and crowdsourced geographical information, including aspects of spatial data quality, patterns of data collection, inequality in coverage and the implications of these patterns to the wider area of Volunteered geographical Information.
The interview is available here .
21 October, 2010
One issue that remained open in the studies on the relevance of Linus’ Law for OpenStreetMap was that the previous studies looked at areas with more than 5 contributors, and the link between the number of users and the quality was not conclusive – although the quality was above 70% for this number of contributors and above it.
Now, as part of writing up the GISRUK 2010 paper for journal publication, we had an opportunity to fill this gap, to some extent. Vyron Antoniou has developed a method to evaluate the positional accuracy on a larger scale than we have done so far. The methodology uses the geometric position of the Ordnance Survey (OS) Meridian 2 road intersections to evaluate positional accuracy. Although Meridian 2 is created by applying a 20-metre generalisation filter to the centrelines of the OS Roads Database, this generalisation process does not affect the positional accuracy of node points and thus their accuracy is the best available. An algorithm was developed for the identification of the correct nodes between the Meridian 2 and OSM, and the average positional error was calculated for each square kilometre in England. With this data, which provides an estimated positional accuracy for an area of over 43,000 square kilometres, it was possible to estimate the contribution that additional users make to the quality of the data.
As can be seen in the chart below, positional accuracy remains fairly level when the number of users is 13 or more – as we have seen in previous studies. On the other hand, up to 13 users, each additional contributor considerably improves the dataset’s quality. In grey you can see the maximum and minimum values, so the area represents the possible range of positional accuracy results. Interestingly, as the number of users increases, positional accuracy seems to settle close to 5m, which is somewhat expected when considering the source of the information – GPS receivers and aerial imagery. However, this is an aspect of the analysis that clearly requires further testing of the algorithm and the datasets.
It is encouraging to see that the results of the analysis are significantly correlated. For the full dataset the correlation is weak (-0.143) but significant at the 0.01 level (2-tailed). However, the average values for each number of contributors (blue line in the graph), the correlation is strong (-0.844) and significant at the 0.01 level (2-talled).
An important caveat is that the number of tiles with more than 10 contributors is fairly small, so that is another aspect that requires further exploration. Moreover, spatial data quality is not just positional accuracy, but also attribute accuracy, completeness, update and other properties. We can expect that they will also exhibit similar behaviour to positional accuracy, but this requires further studies – as always.
However, as this is a large-scale analysis that adds to the evidence from the small-scale studies, it is becoming highly likely that Linus’ Law is affecting the quality of OSM data and possibly of other so-called Volunteered Geographical Information (VGI) sources and there is a decreased gain in terms of positional accuracy when the number of contributors passes about 10 or so.
Completeness in volunteered geographical information – the evolution of OpenStreetMap coverage (2008-2009)
13 August, 2010
The Journal of Spatial Information Science (JOSIS) is a new open access journal in GIScience, edited by Matt Duckham, Jörg-Rüdiger Sack, and Michael Worboys. In addition, the journal adopted an open peer review process, so readers are invited to comment on a paper while it goes through the formal peer review process. So this seem to be the most natural outlet for a new paper that analyses the completeness of OpenStreetMap over 18 months – March 2008 to October 2009. The paper was written in collaboration with Claire Ellul. The abstract of the paper provided below, and you are very welcome to comment on the paper on JOSIS forum that is dedicated to it, where you can also download it.
Abstract: The ability of lay people to collect and share geographical information has increased markedly over the past 5 years as results of the maturation of web and location technologies. This ability has led to a rapid growth in Volunteered Geographical Information (VGI) applications. One of the leading examples of this phenomenon is the OpenStreetMap project, which started in the summer of 2004 in London, England. This paper reports on the development of the project over the period March 2008 to October 2009 by focusing on the completeness of coverage in England. The methodology that is used to evaluate the completeness is comparison of the OpenStreetMap dataset to the Ordnance Survey dataset Meridian 2. The analysis evaluates the coverage in terms of physical coverage (how much area is covered), followed by estimation of the percentage of England population which is covered by completed OpenStreetMap data and finally by using the Index of Deprivation 2007 to gauge socio-economic aspects of OpenStreetMap activity. The analysis shows that within 5 years of project initiation, OpenStreetMap already covers 65% of the area of England, although when details such as street names are taken into consideration, the coverage is closer to 25%. Significantly, this 25% of England’s area covers 45% of its population. There is also a clear bias in data collection practices – more affluent areas and urban locations are better covered than deprived or rural locations. The implications of these outcomes to studies of volunteered geographical information are discussed towards the end of the paper.
4 April, 2010
The opening of Ordnance Survey datasets at the beginning of April 2010 is bound to fundamentally change the way OpenStreetMap (OSM) information is produced in the UK. So just before this major change start to influence OpenStreetMap, it is worth evaluating what has been achieved so far without this data. It is also the time to update the completeness study, as the previous ones were conducted with data from March 2008 and March 2009.
Following the same method that was used in all the previous studies (which is described in details here), the latest version of Meridian 2 from OS OpenData was downloaded and used and compared to OSM data which was downloaded from GeoFabrik. The processing is now streamlined with MapBasic scripts, PostGIS scripts and final processing in Manifold GIS so it is possible to complete the analysis within 2 days. The colour scheme for the map is based on Cynthia Brewer and Mark Harrower‘s ColorBrewer 2.
By the end of March 2010, OpenStreetMap coverage of England grown to 69.8% from 51.2% a year ago. When attribute information is taken into account, the coverage grown to 24.3% from 14.7% a year ago. The chart on the left shows how the coverage progressed over the past 2 years, using the 4 data points that were used for analysis – March 2008, March 2009, October 2009 and March 2010. Notice that in terms of capturing the geometry less than 5% are now significantly under mapped when compared to Meridian 2. Another interesting aspect is the decline in empty cells – that is grid cells that don’t have any feature in Meridian 2 but now have features from OSM appearing in them. So in terms of capturing road information for England, it seems like the goal of capturing the whole country with volunteer effort was within reach, even without the release of Ordnance Survey data.
On the other hand, when attributes are included in the analysis, the picture is very different.
The progression of coverage is far from complete, and although the area that is empty of features that include street or road name in Meridian 2 is much larger, the progress of OSM mappers in completing the information is much slower. While the geometry coverage gone up by 18.6% over the past year, less than 10% (9.6% to be precise) were covered when attributes are taken into account. The reason for this is likely to be the need to carry a ground survey to find the street name without using other copyrighted sources.
The attribute area is the one that I would expect will show the benefits of Ordnance Survey data release to OSM mapping. Products such as StreetView and VectorMap District can be used to either copy the street name (StreetView) or write an algorithm that will copy the street name and other attributes from a vector data set – such as Meridian 2 or VectorMap District.
Of course, this is a failure of the ‘crowd’ in the sense that as this bit of information previously required an actual visit on the ground and it was a more challenging task than finding the people who are happy to volunteer their time to digitise maps.
As in the previous cases, there are local variations, and the geography of the coverage is interesting. The information includes 4 time points, so the most appropriate visualisation is one that allows for comparison and transition between maps. Below is a presentation (you can download it from SlideShare) that provides maps for the whole of England as well as 5 regional maps, roughly covering the South West, London, Birmingham and the Midlands, Manchester and Liverpool, and Newcastle upon Tyne and the North West.
If you want to create your own visualisation, of use the results of this study, you can download the results in a shapefile format from here.
For a very nice visualisation of Meridian 2 and OpenStreetMap data – see Ollie O’Brien SupraGeography blog .
29 January, 2010
After the publication of the comparison of OpenStreetMap and Google Map Maker coverage of Haiti, Nicolas Chavent from the Humanitarian OpenStreetMap Team contacted me and turned my attention to the UN Stabilization Mission in Haiti’s (known as MINUSTAH) geographical dataset, which is seen as the core set for the post earthquake humanitarian effort, and therefore a comparison with this dataset might be helpful, too. The comparison of the two Volunteered Geographical Information (VGI) datasets of OpenStreetMap and Google Map Maker with this core dataset also exposed an aspect of the usability of geographical information in emergency situations that is worth commenting on.
For the purpose of the comparison, I downloaded two datasets from GeoCommons – the detailed maps of Port-au-Prince and the Haiti road network. Both are reported on GeoCommons as originating from MINUSTAH. I combined them together, and then carried out the comparison. As in the previous case, the comparison focused only on the length of the roads, with the hypothesis that, if there is a significant difference in the length of the road at a given grid square, it is likely that the longer dataset is more complete. The other comparisons between established and VGI datasets give ground to this hypothesis, although caution must be applied when the differences are small. The following maps show the differences between the MINUSTAH dataset and OpenStreetMap and MINUSTAH and Google Map Maker datasets. I have also reproduced the original map that compares OpenStreetMap and Map Maker for the purpose of comparison and consistency, as well as for cartographic quality.
The maps show that MINUSTAH does provide fairly comprehensive coverage across Haiti (as expected) and that the volunteered efforts of OpenStreetMap and Map Maker provide further details in urban areas. There are areas that are only covered by one of the datasets, so they all have value.
The final comparison uses the 3 datasets together, with the same criteria as in the previous map – the dataset with the longest length of roads is the one that is considered the most complete.
It is interesting to note the south/north divide between OpenStreetMap and Google Map Maker, with Google Map Maker providing more details in the north, and OpenStreetMap in the south (closer to the earthquake epicentre). When compared over the areas in which there is at least 100 metres of coverage of MINUSTAH, OpenStreetMap is, overall, 64.4% complete, while Map Maker is 41.2% complete. Map Maker is covering further 354 square kilometres which are not covered by MINUSTAH or OpenStreetMap, and OpneStreetMap is covering further 1044 square kilometres that are missing from the other datasets, so clearly there is a benefit in integrating them. The grid that includes the analysis of the integrated datasets in shapefile format is available here, in case that it is of any use or if you like to carry out further analysis and or visualise it.
While working on this comparison, it was interesting to explore the data fields in the MINUSTAH dataset, with some of them included to provide operational information, such as road condition, length of time that it takes to travel through it, etc. These are the hallmarks of practical and operational geographical information, with details that are relevant directly to the end-users in their daily tasks. The other two datasets have been standardised for universal coverage and delivery, and this is apparent in their internal data structure. Google Map Maker schema is closer to traditional geographical information products in field names and semantics, exposing the internal engineering of the system – for example, including a country code, which is clearly meaningless in a case where you are downloading one country! OpenStreetMap (as provided by either CloudMade or GeoFabrik) keeps with the simplicity mantra and is fairly basic. Yet, the scheme is the same in Haiti as in England or any other place. So just like Google, it takes a system view of the data and its delivery.
This means that, from an end-user perspective, while these VGI data sources were produced in a radically different way to traditional GI products, their delivery is similar to the way in which traditional products were delivered, burdening the user with the need to understand the semantics of the different fields before using the data.
In emergency situations, this is likely to present an additional hurdle for the use of any data, as it is not enough to provide the data for download through GeoCommons, GeoFabrik or Google – it is how it is going to be used that matters. Notice that the maps tell a story in which an end-user who wants to have full coverage of Haiti has to combine three datasets, so the semantic interpretation can be an issue for such a user.
So what should a user-centred design of GI for an emergency situation look like? The general answer is ‘find the core dataset that is used by the first responders, and adapt your data to this standard’. In the case of Haiti, I would suggest that the MINUSTAH dataset is a template for such a thing. It is more likely to find users of GI on the ground who are already exposed to the core dataset and familiar with it. The fields are relevant and operational and show that this is more ‘user-centred’ than the other two. Therefore, it would be beneficial for VGI providers who want to help in an emergency situation to ensure that their data comply to the local de facto standard, which is the dataset being used on the ground, and bring their schema to fit it.
Of course, this is what GI ontologies are for, to allow for semantic interoperability. The issue with them is that they add at least two steps – define the ontology and figure out the process to translate the dataset that you have acquired to the required format. Therefore, this is something that should be done by data providers, not by end-users when they are dealing with the real situation on the ground. They have more important things to do than to find a knowledge engineer that can understand semantic interoperability…
As the relief effort to the crisis in Haiti unfolds, so does response from mapping organisations with global reach. It is a positive development that free data is available from the Volunteered Geographic Information (VGI) community to assist humanitarian work on such a large scale, and good that there are now two sources. However, it is sad to discover that there seems to be friction between Google Map Maker and OpenStreetMap as to which organisation will prevail among governmental and NGO users. A key issue is surely to ascertain – and fast - which source of crowdsourced geographic information is most useful for which geographical area, and where the differences lie.
I did this assessment today, in the hope that it is useful for the emergency relief work now, and for the reconstruction work to follow. The data is current for the 18th January 2010, and the results are available here.
The evaluation of the coverage of Google Map Maker and OpenStreetMap for Haiti was done using the same methodology as for the comparison of OpenStreetMap and Ordnance Survey data. The shapefile’s projection is UTM zone 18N. In the map here, yellow means that there is a better coverage in Map Maker, and blue means that there is a better coverage in OpenStreetMap. The difference between the two datasets is expressed in metres.
Unlike the previous comparison, where it was assumed that one dataset was the more accurate, here it is not helpful to pursue a binary approach. Rather, there are differences between the two sources of data, and these may matter as the relief work is carried out. The evaluation question is: for each grid square, which of the datasets contains more information in terms of roads length?
The file contains the total roads length for both datasets. The calculated difference between them using the equation:
∑(OSM roads length)-∑(Map Maker roads length)
for each 1km grid square.
The information in the file can be used for the following applications:
- Users of these mapping products - it can help in judging which dataset to use for each area.
- Users – it can facilitate conflation - the process of merging datasets to create a better quality output.
- Mappers - it can illuminate which areas to focus on, to improve coverage.
If you download the file, notice that the field OSMMMClose indicates that the two datasets are very close to one another – the value 1 is associated with grid squares where the difference between them is less than 200 metres. This might be useful as an indication that the two datasets agree with each other.
I hope that this assessment is helpful for those using the data for the relief effort. If you have ideas on how I can help further in this way, please get in touch.