Monte Carlo simulation as a methodology for coping with error in Geographic Information Systems.

摘要

Error has always been and will always be a problem in geographic analysis. Before automation, this problem was handled by highly skilled individuals who were involved in every step of the analysis process and knew intuitively how far the results could be relied upon. However, with today's automation, data and operational error problems are frequently transparent to the operator.;In concept, the problem of error propagation in Geographic Information Systems (GIS) can be handled using existing statistical theory. However, in practice GIS processes are too complex for such an approach. The Monte Carlo method has been suggested by a few in the geographic modelling community as an immediately available way of dealing with the problem in GIS. The method, which has been considered virtually a universal method for problems that could not be solved by other means, has not been widely used in GIS because of the increased computational load and a belief in more traditional analytic methods of problem solving.;My purpose was to determine whether a Monte Carlo simulation could provide the current GIS user with a useful approach to error prediction. To temper the idealistic conditions of a laboratory environment, I took the approach of a case study examining a bear-bones system currently in wide use by the U.S. Army. In this context, three claims were investigated: (1) error propagation in GIS can be a problem, (2) learning to cope with error is realistic for GIS users, and (3) the method is an acceptable technique.;The results indicate that GIS error can be a particular problem when the machine's results are applied outside the realm of situated human judgement. In this respect, all methods of predicting error are simply tools and learning to cope with error is necessary for the user. Despite the obvious drawback of the increased computational load associated with the Monte Carlo method, simulation is the only workable tool for predicting error in GIS results.