ERROR PROPAGATION MODELING IN RASTER GIS - OVERLAY OPERATIONS

Performing data manipulations on maps that possess error as a result o f the process of data collection leads to error propagation. The error s that are present in maps are modified by such operations in ways tha t may undermine the purpose of analysis and lead to increased uncertai nty in the validity of the conclusions that are drawn. This paper anal yses how source map error propagates as a result of overlay operations . Geman and Geman's corruption model for individual source map error i s used for the analysis which allows for attribute measurement error a nd location error that can then interact with the (true) source map ge ography. This paper reports theoretical results on the univariate over lay problem and then extends these results through simulation. Through out a set of source maps and error processes are used with specified p roperties in order to examine in detail the interactions that can take place between the different elements of the source map structure and the error process. The paper uses ANOVA methods to quantify the contri bution made by different components of the map and the measurement pro cess to spatial (join count) and non-spatial (proportion of misclassif ied pixels and the Kappa index) descriptors of error. The paper conclu des with a discussion of the usefulness of these results for managing error in spatial databases and for the development of automated uncert ainty reporting.