PROBABILISTIC METHODS FOR ASSESSING RELIABILITY OF CONTAMINATED SEDIMENT VOLUME ESTIMATES

Abstract

Sampling requirements for the quality control of environmental sediment contamination characterization are not currently explicit. The effect of sample size on the accuracy of estimated volume and variability of contaminated sediments is important to quantify, as these factors potentially have major impacts on the choice of remedial action. Random field simulation has been shown to be an effective method of representing sediment thickness variation over space, and such models have been effective in the risk assessment of similar environmental studies. In this thesis, the Kriging geostatistical model, and random field simulative model, local average subdivision (LAS), have been applied to an ongoing remediation project in Boat Harbour, Nova Scotia, Canada. The objective of this thesis is to assess the effect of sample size on the accuracy of volume estimation and its variability when comparing the more common geostatistical modeling methods versus the more novel, and promising, random field simulative methods. This thesis compares the two modeling methods at various sample sizes and discusses further implications of model effectiveness for remediation practitioners. It is found that the Kriging and LAS models produce similar results at higher sample densities (i.e. those exceeding 2.6 samples/ha), however the LAS model produces higher precision and accuracy in its estimate of the total sediment volume. Furthermore, it is concluded that, for remedial practitioners, LAS is the more effective and conservative modeling technique in comparison to Kriging for sediment volume characterization.