Effects of the loss of correlation structure on Phase 1 dose estimates. Hanford Environmental Dose Reconstruction Project.

摘要

In Phase I of the Hanford Environmental Dose Reconstruction Project, a step-by-step (modular) calculational structure was used. This structure was intended (1) to simplify the computational process, (2) to allow storage of intermediate calculations for later analyses, and (3) to guide the collection of data by presenting understandable structures for its use. The implementation of this modular structure resulted in the loss of correlation among inputs and outputs of the code, resulting in less accurate dose estimates than anticipated. The study documented in this report investigated two types of correlations in the Phase I model: temporal and pathway. Temporal correlations occur in the simulation when, in the calculation, data estimated for a previous time are used in a subsequent calculation. If the various portions of the calculation do not use the same realization of the earlier estimate, they are no longer correlated with respect to time. Similarly, spatial correlations occur in a simulation when, in the calculation, data estimated for a particular location are used in estimates for other locations. If the various calculations do not use the same value for the original location, they are no longer correlated with respect to location. The loss of the correlation structure in the Phase I code resulted in dose estimates that are biased. It is recommended that the air pathway dose model be restructured and the intermediate histograms eliminated. While the restructured code may still contain distinct modules, all input parameters to each module and all out put from each module should be retained in a database such that subsequent modules can access all the information necessary to retain the correlation structure.