Do different methods of representing real-world features have an effect on the findings from GIS analyses? To what extent does choice of data sources affect network analysis? In considering network accessibility, are results affected by the representation of supply and demand considerations? 1 Copyright © by the paper's authors. Copying permitted for private and academic purposes. In: A. Comber, B. Bucher, S. Ivanovic (eds.): Proceedings of the 3rd AGILE PhD School, Champs sur Marne, France, 15-17-September-2015, published at http://ceur-ws.org
The number and variety of sources of spatial data continues to expand, as do the debates regarding the quality and usability of such data, particularly those which are Free and Open Source (FOS) or free-to-use. The highest quality data is often expensive to obtain and the option of cost-free data sets is tempting for many users.
With the existence of the huge hinterland of data quality research acknowledged,
and a great number of studies investigating the usability of devices and interfaces,
little attention has been paid to the usability of data, and even less into the
usability of geographical data in typical GIS research situations. There has been some
research into the use of volunteered geographic information (VGI) in the field of
data quality theory and assessment
This study set out to address cross-cutting themes that are topical in GIS and geographical analysis given trends towards the use of open source data, namely: There is little evidence to date on which to quantify the effects of these issues on final results. This research is intended to take a step in redressing gaps that exist in the knowledge, understanding and perception of such data.
This study argues that even the best quality data may not be appropriate in certain contexts. To highlight the type of scenarios where this may indeed be the case several commercial and free-to-use data sources were used in sensitivity analyses of the application of well-established GIS network analysis tasks. The aim is to assess whether findings vary according to the application of alternate data sets used to represent the same features within such models.
The research took the form of various case studies, all tied around a similar theme, that of accessibility. Some of the studies assessed accessibility to features that have been the subject of much research in the past (such as GP surgeries), while some looked at less commonly assessed features (such as primary schools, secondary schools and sports facilities). All were linked by an interest in various health and fitness initiatives and investigations that have taken place in South Wales (UK), such as those looking at active travel to schools, equitable access to health care and reasonable geographical access to sport and leisure facilities. The part of the study relating to accessibility to primary schools will be used as an example.
The supply feature (primary schools) were represented in four different ways by the two datasets examined: a Point of Interest1 point (nominally the centroid of the main school building); the pedestrian access points of each school; the geometric centroid of the entire school site (including play areas, sports fields and car parks); and the site perimeter. The Ordnance Survey Sites dataset2, by providing the footprint of each school as well as the access points, offered more detail and precision to measurements of access, raising another interesting question as to whether any increase in precision automatically resulted in an increased accuracy of results. The places of origin for journeys to the schools were kept constant, and were UK census Output Area population-weighted centroids (the smallest unit of published UK census data).
Distances from each population centroid were measured to the nearest school, looking at each representation in turn, using the various network datasets. The network datasets included commercial data (Ordnance Survey ITN and ITN with Urban Paths3), free-to-use data (Ordnance Survey OpenRoads4) and FOS data 1 https://www.ordnancesurvey.co.uk/business-and-government/products/pointsof-interest.html
2https://www.ordnancesurvey.co.uk/business-andgovernment/products/topography-layer.html
3https://www.ordnancesurvey.co.uk/business-and-government/products/itnlayer.html
(OpenStreetMap5). Sensitivity analysis was conducted through repetitions of the
distance calculations, ensuring every combination of network (plus Euclidean
measurement) was used for every feature representation. The process was then
repeated in its entirety using a Two-Step Floating Catchment Area (2SFCA)
measurement. As described by
The large number of results generated were cross-compared. The comparison revealed that for primary schools the vast majority of results (over 80% of all comparisons) were statistically significantly different from the others at the < .001 level, for both distance and 2SFCA measures. This indicated that the different datasets used were not interchangeable and therefore not equally usable in this type of study.
At this early stage of analysis initial indications were that differences between the network datasets had the greater effect on results. Differences due to method of demand- or supply-side feature representation were less important. Initial findings suggest that more attention needs to be given to the nature of data sets used to represent such features in GIS-based analytical tasks. The exact context in which such data sets are applied may determine how usable different sources of data are in relation to common GIS spatial analytical tasks and a useful addition to GIS-based analysis going forward could be the derivation of a typology of circumstances in which adopting alternative sources of open data are more appropriate.
Acknowledgments This research was funded by Ordnance Survey (OS) but any interpretations of findings are those of the student and do not necessarily reflect the opinions of OS. 4https://www.ordnancesurvey.co.uk/business-and-government/products/os-openroads.html 5http://www.openstreetmap.org/