Calibration of Multimedia Models Using Bayesian Techniques: Results of International Model-Data Intercomparisons. I. Linkov, ICF Consulting; and D. Burmistrov, Menzie-Cura & Associates
Bayesian updating has been found to be a useful methodology that gives the analyst the ability to incorporate a limited knowledge of a specific ecosystem in order to improve model predictions and reduce or reconcile associated uncertainties. Our experience shows that Bayesian techniques are especially valuable for the calibration of complex multimedia model where models are expected to describe large and often contradictory datasets. This study illustrates the application of Bayesian calibration techniques for two model-data intercomparison scenarios developed by the International Atomic Energy Agency (IAEA) Fruit and Forest Working Groups under the BIOMASS program. FORESTPATH and FRUITPATH are probabilistic multimedia models that were developed for the prediction of public and environmental dose and risk assessments associated with environmental contamination by radionuclides and heavy metals. Bayesian updating is proposed as a tool to reduce uncertainty associated with model predictions and to tailor generic models to specific sites. The preliminary results obtained for the two model-data scenarios have shown that there is generally a high degree of consistency between predictions made by calibrated models and experimentally-measured data. Difficulties associated with application of Bayesian techniques will be also discussed.
This work was done as part of the IAEA BIOMASS Program.
Go to . . .
2002 SRA Annual Meeting Table of Contents
2002 SRA Annual Meeting Author Index
Main Abstracts Menu Page
RiskWorld Home Page