UNCERTAINTY EVALUATION OF MILITARY TERRAIN ANALYSIS RESULTS BY SIMULATION AND VISUALIZATION
Pekka Horttanainen, Kirsi Virrantaus · 2004
This article introduces our research on the reliability of military terrain analysis results by using a spatially dependent uncertainty model and stochastic simulation. Soil map is the most important source data set in the analysis. Geologists have produced soil maps by using manual evaluation procedures and no metadata are automatically included. On the basis of the expert knowledge of geologists, a misclassification matrix can be created. A spatial autoregressive stochastic process has been applied using probability information derived from the misclassification matrix. A parameter that shows the level of the spatial dependency is the crux of this method and we have analysed it as well as the other results of the analysis by using manual evaluation procedure based on visualization. Our uncertainty model is quite simple and in our future work we want to increase the use of expert knowledge. It can occur basically in two ways: first by replacing the probability information of the misclassification matrix by possibility knowledge derived from fuzzy soil map probably supported by kriging and secondly by developing the use of the parameter of spatial dependency to describe better different soil types.