A statistical model of possible future human actions

A critical review has been carried out of models of possible future human actions during the long-term post-closure period of a radioactive waste repository. Various Markov models have been considered as alternatives to the more conventional Poisson model, and the problems of parameterisation have been addressed. Where the Poisson model unduly exaggerates the intrusion risk, some form of Markov model may have to be introduced. This situation may well arise for shallow repositories, but it is thought less likely for deep repositories. Consideration of state transition within a Markov model highlights the value of increasing the availability of site information. (author)