Towards Quality-Aware GIS: Operation-based retrieval of spatial data quality information

Despite the fact that all geospatial data are, at different levels, imperfect, GIS technologies do not offer much support to GIS users to make them aware of the imperfections and potential limitations in the use of these data. Recent research progresses have been made towards the design of GIS that would take into account data quality. These approaches extend typical GIS functionalities to manage, use and communicate information describing spatial data quality. This paper presents a project that aims at linking quality information with GIS operations to provide users with the most relevant quality information. Background and Relevance All geospatial data are at different levels imperfect (Goodchild, 1995, Devillers and Jeansoulin, 2006). This results from a number of factors, such as human errors in the recording of the data or interpretation of the world features, errors arising from the manipulation of the data, representation of complex phenomena with an inappropriate data model, imprecision of the equipment used to measure the phenomena, inappropriate sampling, etc. To record this information and release it to GIS users, a number of geospatial data producers provide metadata (i.e. data about data) that describe some aspects of the data quality of the datasets. Such metadata are through to be helpful to the users for assessing the fitness of the data for their specific usage. Standardization bodies, such as the Federal Geographic Data Committee (FGDC) in the USA or the International Standard Organisation (ISO) internationally, provide standards for documenting such metadata. These metadata are however criticized by a number of GIS experts for different reasons. They are often too technical to be understood by the intended audience (Timpf et al., 2006, Boin and Hunter, 2006), they do not describe all the types of data quality issues (Comber, 2007), they are usually too general (e.g. describe the dataset as a whole and not individual objects) (Devillers, 2005, Sadiq, 2008), they are rarely linked properly to the data they describe, etc. In practice, metadata end-up to be ignored or not understood by the majority of GIS users. In addition, to encourage their adoption, metadata standards request only few metadata to be documented (ISO TC/211, 2003). These minimal metadata are often referred as “discovery metadata” and typically include, for instance, the spatial and the temporal extents of the dataset and the dataset title and language, but exclude all aspects of data quality information. As a consequence, most metadata files provide no information about data quality at all and GIS users are left without any information about the quality of the data they are using. Some commercial GIS (e.g. ArcGIS) now provide the ability to manage, update and display metadata. However, these metadata are hard to find for users with little expertise in GIS and also lack of any real connection with GIS operations. Metadata are stored but do not affect the functioning of the GIS. For instance, the result of a GIS operation (e.g. a distance measurement between two objects on a map) will not change if the user has data of very good quality or data of very poor quality. In a context where geospatial information gets increasingly used by the general public, the fact that GIS work as if they were handling perfect data can result in poor decisions being made. For these reasons, a number of research projects have been directed towards improving GIS to have them provide additional functionalities related to spatial data quality. Some work focused on improving the management of data quality information and allowing, for instance, its storage using different database models and at various levels of details (e.g. Devillers et al. 2005, Sadiq, 2008). Other works looked at communicating quality information to users through different visualization techniques, such as thematic maps of certain aspects of data quality (e.g. see Beard and Buttenfield, 1999 for an overview of such approaches), display of quality indicators (Devillers et al. 2005, Huth et al. 2007), display of warning messages (Reinke and Hunter, 2002, Levesque et al., 2007). This paper focuses on an ongoing project that looks at relating the data quality information stored in the metadata to GIS operations.