A geostatistical approach to modelling positional errors in vector data

As part of the theoretical development and practical applications of GISs, error issues are receiving increasing attention. This paper contributes to the debate in GIS error issues by exploring the applications of geostatistics in vector data, where positional errors are of major concern. A review is provided of the methods for handling positional errors in GIS vector data comprising points and lines. This is followed by a description of a stochastic simulation approach to modelling positional errors, which is remarkable for its ability to accommodate the spatial correlation characteristics to spatial data and their errors. Results from an experiment using photogrammetric data confirm the effectiveness of the proposed approach for modelling positional errors. The simulation approach is also examined with respect to other methods where due consideration is not given to the spatial correlation that is intrinsic to positional errors. Stochastic simulation-based modelling of uncertain vector data via raster structures represents a valuable extension and contribution of geostatistical approaches to integrated handling of errors in heterogeneous spatial data.