An outline of the uncertainty handling and data assimilation framework
within a Bayesian frame is given, illustrated by its use within RODOS
. Particular emphasis is paid to the need for compatible methodologies
and data structures to hold uncertainty assessments throughout all th
e modules of a decision support system. Progress at developing techniq
ues and modules is described: (i) to use belief nets to predict the so
urce term when an accidental release threatens; (ii) to estimate the s
ource term using a ring of gamma monitors at the periphery of the plan
t; and (iii) to estimate the source term from both near and more dista
nt measurements. The management of uncertainties in the food chain mod
elling is briefly indicated. Finally, the issue of moving the basis of
prediction from modelling approaches to databases of environmental me
asurements is discussed.