Bayesian Model Development for Analysis of Open Source Information to Support the Assessment of Nuclear Programs

Pacific Northwest National Laboratory has spent several years researching, developing, and validating large Bayesian network models to support integration of open source datasets for nuclear proliferation research. Our current work focuses on generating a set of interrelated models for multi - source assessment of nuclear programs, in contrast to a single c omprehensive model . By using this approach, we can break down the models to cover logical subproblems that can utilize different expertise and data source s . This approach allows researchers to utilize the models individually or in combination to detect and characterize a nuclear program and identify data gaps . The models operate a t various levels of granularity, covering a combination of state - level assessme nts with more detailed models of site or facility characteristics. This paper will describe the current open - source - driven, nuclear nonproliferation models under development, the pros and cons of the analytical approach, and areas for additional research .