Impact of Accuracy of Prior Information and Geophysical Sensors on Geostatistical Characterization of UXO Sites

Category: Site Characterization Presented at: The UXO/Countermine Forum, September 3-6, 2002, Orlando, Florida ABSTRACT There are different sources of uncertainty attached to UXO site characterization. One of them is associated with prior information regarding locations of ordnance targets, which are used initially for designing sampling schemes. The accuracy of information is then defined by the spatial bias in target coordinates. The other uncertainty arises during the preparation of geophysical sensors. Their quality can be summarized by a pair of probabilities: 1) the tolerable probability of false alarm and 2) the probability of detecting UXO calculated directly from the selected false alarm using ROC curves. This paper examined the impact of the quality of prior information available and the choice of an acceptable false alarm rate on geostatistical characterization of unexploded ordnance (UXO) sites. The risk of occurrence of at least one UXO is mapped using geostatistics for three different levels of spatial bias for prior information and three different probabilities of false alarm. Once risks are mapped, pixels are flagged for further investigation if the estimated probabilities of occurrence exceed a given threshold. This classification is used to compare the performance of different scenarios. Classification results indicate that accurate prior information keeps the false negative rate low. Also perfect sensors are not necessarily cost effective because of preferential sampling problems.