Context

In the past, only formal data sources were available for spatial analysis. Whilst the quality of this formal data was high, there were limitations in its use and application due to costs and licencing restrictions. Some ‘open’ formal data has become available for all to use for free under government initiatives, but the majority of formal data still has very clearly defined, restrictive and potentially costly licencing agreements. Many users of geographic data have sought to find a ‘work around’ or other options to avoid these restrictions. As a result there has been a massive growth in the development of, contribution to and use of informal data sources in the last decade.

The rise of informal data sources or Volunteered Geographic Information (VGI) (Goodchild, 2007) has been phenomenal. There is now a range of web based projects, such as OpenStreetMap (OSM) (OpenStreetMap, 2014) and Wikimapia (Wikimapia, 2014) that allow user communities to create, manage and share geographic datasets all with little or no cost to subsequent users.

However, the quality and completeness of polygonal features within OpenStreetMap (OSM) still remains to be investigated, measured and quantified. As such, this research looks to analyse the data quality of the informal (OSM) polygonal data against a reference polygonal dataset from Ordnance Survey (OS). To achieve this goal, the research looked at schools, as they play a significant part of public life. By focusing on school data (as a manageable subset) it gives a useful indicator to illustrate the wider quality of polygonal data in OSM.

Methodology and datasets

Due to the range of the data and research undertaken, the analysis has been separated into four clear areas of activity each using a relevant methodology (Figure 1).

Methodology 1 - point to point analysis of the government and council datasets.
Methodology 2 - completeness of the OS polygons compared to the council point datasets.
Methodology 3 - analysis of the completeness and accuracy between the OS polygons and the OSM polygons.
Methodology 4 – a review of the attribution and metadata associated with OSM and the other formal datasets.

Figure 1 – Methodologies and outputs

For this research project, formal data was gathered from local council authorities, the Scottish Government (SG) and the national mapping agency, Ordnance Survey (OS). The quality of informal data is known to vary according to positional accuracy, attribute accuracy and geographic region (urban v’s rural). Therefore, three local council authority areas within Scotland were used as case study areas. The aim was to complete a comparison between study areas across a range of different geographic locations and authority sizes. These were the City of Edinburgh Council (CEC) an urban city authority, Perth and Kinross Council (PKC), a rural authority which includes the city of Perth and finally the Scottish Borders Council (SBC) covering a mainly rural landscape with a number of medium sized towns (Figure 2).

Figure 2 – Local council case study areas

Key results

Methodology 1 – geocoded postcodes cannot be relied upon to pin-point specific entities such as schools (Figure 3).

Figure 3: Distance of over 2km between geocoded postcode and actual school (maximum distance found)

Methodology 2 – It should not be assumed that an OS reference dataset is complete (Figure 4).

Figure 4: Panmure St Ann’s school does not appear on the OS dataset

Methodology 3 – Polygonal overlap between OSM and the OS reference dataset varies greatly, but is higher in rural areas (Figure 5 and Figure 6).

Figure 5: 90% overlap between OS and OSM datasets

Figure 6: 40% overlap between OS and OSM datasets

Methodology 4 – OSM attribution is very patchy and incomplete. It seems that OSM contributors like to draw objects but are not interested/don’t understand the benefits of good attribution.

Interpretation of results

With methodology 1, we looked at council data against Scottish Government (SG) data. What was clear was that while the SG data was very up-to-date, the geocoding of the postcodes had limited value other than giving a general point in the general vicinity of where a school was located. It also highlighted that while councils have very accurate data, it seemed that there is no regular schedule for data review (as the Scottish Government has) and as a result old and new schools can be missing from the relevant datasets.

In methodology 2, we investigated the quality and completeness of the reference dataset provided by OS. It was proven to be very high, but not 100% complete. Issues flagged were mostly of a temporal nature and occurred in the local council data as well as the OS reference dataset.

Methodology 3 compared OpenStreetMap (OSM) and Ordnance Survey (OS) polygonal data. It was clear from the outset that the OSM dataset was incomplete compared to the OS reference dataset. However, it was interesting to see how the data compared across the different urban and rural areas. There was a greater polygon overlap/accuracy in rural areas than in urban areas. As such it can no longer be assumed that a greater number of contributors focused on urban area will create the most accurate data. It comes down to the enthusiasm of the individual contributors, patience and/or local knowledge which makes them very thorough in the mapping activities that they complete.

In Methodology 4 there was an examination of the attribution options and quality across the formal and informal datasets. The lack of attribution across the OSM polygonal data significantly weakened the overall effectiveness of the data gathered and as a result, limits how it could be used.

Conclusions

The aim of this research was to investigate the quality and completeness of the spatial data for schools across three case study areas in Scotland. It examined and compared different point and polygon datasets from a range of sources (local councils, the Scottish Government, OS and OSM) to ascertain what levels of accuracy and completeness existed and to measure VGI’s fitness for use.

The results showed that the OSM polygonal data investigated, does not have sufficient geometric completeness or accuracy to be used instead of the OS reference dataset. In addition, the OSM attribution on the polygons (that were examined) was very incomplete and as a result further weakened the opportunities for it use in spatial analysis applications.

Having completed this polygonal based research and comparing and contrasting it to previous linear research (Haklay, 2010), it has become clear that measuring the ‘fitness for use’ of OSM data needs to be considered on a feature by feature basis (Table 1).

Table 1: OSM feature types, their level maturity and fitness for use

OSM Feature type	Maturity and viability for commercial use
Linear	The linear contributions in OSM (roads) have reached a point of completeness and maturity where they can be effectively used for meaningful networking analysis and way finding applications.
Polygonal	The polygonal contributions in OSM (like the schools subset examined in this paper) provide a cartographic/visual representation for users, but have not yet reached a point where they are robust or consistent enough for spatial analysis.

References

GOODCHILD, M.F. 2007. Citizens as voluntary sensors: Spatial data infrastructure in the world of

web 2.0. Int. J. Spat. Data Infrastr. Res., 2, 24–32.

HAKLAY, M. 2010. How good is volunteered geographical information? A comparative study of OpenStreetMap and Ordnance Survey datasets. Environment and Planning B-Planning & Design, 37, 682-703.

OPENSTREETMAP. 2014. OpenStreetMap.

[ONLINE] Available at: http://www.openstreetmap.org/.

[Accessed 30 July 2014].

WIKIMAPIA. 2014. Wikimapia - Let's describe the whole world!.

[ONLINE] Available at: http://wikimapia.org/.

[Accessed 30 July 2014].