Past weapons production activities have resulted in mass quantities of trans-uranic waste being buried in drums at several sites in the United States. In an effort to relocate these waste drums to more permanent storage sites, Fluor Hanford has begun characterizing their contents to ensure compliance with various shipping and storage requirements. Non-destructive analysis techniques are regularly employed, among them passive radiation detection using a Canberra Gamma-Energy-Analyzer germanium detector vault. Necessary strict legal tolerances require strong quality assurance. The detectors are frequently calibrated in the traditional method with check sources, but it would be advantageous to have an estimate of system minimum detectable activity (MDA). However, any estimate is complicated by the fact that sources are distributed stochastically in the waste drums. In this study, a method was developed to predict system detector efficiency for a variety of detector configurations and drum fill materials and calculate MDA based on these efficiencies. The various system designs were modeled in Monte Carlo N-Particle Code, version 4b, to determine photopeak detection efficiency. An external code written in C programming language was used to randomly assign between one and 20 sources to volumetric regions of the waste drum. Twenty simulations were performed for each design and drum fill material combination, each time redefining the stochastically distributed source. This provided a normally distributed spectrum of 20 efficiencies for each situation. From this, mean and lower 95% confidence limit efficiencies were used to calculate MDA. The patterns among the results were then compared with values predicted by the MDA formula. Finally, an examination was made of the impact on the MDA of the system's true design in the case of single or multiple detector failure. The results indicate that this method of estimating minimum detectable activity, although costly in computing time, provides results consistent with intuitive and calculated expectations. Future work would allow easy calibration of the model to measured efficiency results. Used in coordination with physical experiments, this method may eventually prove useful in benchmarking system performance and accurately ensuring reliable waste drum characterizations. / Graduation date: 2003