Uncertainties in modeling groundwater contamination
Stockholm University2003 (English)Doctoral thesis, comprehensive summary (Other academic)
The scientific problem of nuclear waste disposal suffers from strong
uncertainties arising both from its object (the system being studied being unknown) and also from its ambitious objectives (prediction of future events and their consequences). This work focuses on identifying, structuring and quantifying the main uncertainties involved in this field. It also shows the way how to integrate all the identified uncertainties into any inference and prediction a risk analyst might want to make in this field. An overall conceptual support is provided by using Bayesian statistics combined with hierarchical modeling techniques. Scenario and parametric uncertainty and their implications were studied in the context of a expert assessment project called GESAMAC, where Functional Data Analysis techniques were used to study the variability of the dose vs. time curves. Markov Chain Monte Carlo simulation was also used to make inference studies of parametric uncertainty using data from a field study.
Although this study is not exhaustive, it does show that the different sources of uncertainties do have an important impact on what we really know about the development system. Acknowledging scenario uncertainty, for instance, shows that we should expect uncertainties of dose responses on the order of 20 times what is expected to happen by default. It is shown that there is an asymmetry on the uncertainty of the dose curves; it seems there is more uncertainty on how the curve reaches its maximum than on how it relaxes afterwards. It is also found that the influence of the physical parameters on the dose curve is far from being linear and it seems to depend in a very important way on large scale physical inhomogeneities of the system.
Place, publisher, year, edition, pages
Stockholm: Department of Physics, Stockholm University , 2003. , 42 p.
Research subject Physics
IdentifiersURN: urn:nbn:se:su:diva-7819ISBN: 91-7265-574-7OAI: oai:DiVA.org:su-7819DiVA: diva2:199098
2003-01-28, FB42, Albanova, Roslagstullsbacken 21, Stockholm, 10:00 (English)
Ekberg, Christian, Docent