View Report Details

Probabilistic modelling for the aquatic environment

Project Code: R01036

31/03/2001

Centre for Environment, Fisheries & Agriculture Science (CEFAS) Lowestoft
Grzechnik, M ; Round, G; Brownless, G; Camplin, W

Under contract SA168, the Food Standards Agency asked CEFAS to develop a probabilistic model for predicting dose distributions from radionuclides discharged to the marine environment.

CEFAS’ deterministic models for dose assessments have been developed to allow the model parameter uncertainties to be propagated through the codes. A probability density function of water concentrations is output from WAT (renamed as WATP for probabilistic modelling), which is fed into ADO (ADOP) which then outputs a probability density function of doses. The new codes were implemented in a Windows PC environment, using FORTRAN 77.

Two different methods of selecting random numbers from specified distributions for use in the model were evaluated for accuracy (as the efficiency of each was similar enough to not be an issue). These were Simple Random Sampling and Latin Hypercube Sampling. The latter was deemed to be the best choice as it proved to give smoother distribution predictions for the input considered.

The uncertain parameters were identified and the uncertainty distributions (for the entire UK coast) determined by informal expert elicitation of CEFAS Scientists. These distributions were then propagated through each of the models using the technique known as Monte Carlo analysis. This results in distributions of water concentrations (WATP) and doses (ADOP), which have been plotted and analysed using Microsoft Excel.

A case study was conducted whereby the new models were applied to an assessment at Sizewell to investigate the issues that may arise in their regular usage. In this case, distributions of critical group consumption and occupancy/exposure rates have been defined according to the results of habits surveys. The outputs have been compared to those obtained with a deterministic assessment, and the uncertainty due to the habits survey has been investigated.

The main conclusions drawn are:

The habits survey should be included in all probabilistic assessments as a distribution of values. Whether this distribution should take into account the variability of the entire habits survey group or the highest users only is the subject of further work to be considered (the former approach is taken here).

percentile of the dose output distribution is recommended for use as an indicator of the upper bound of dose to the critical group. This value is usually always higher than that obtained using the deterministic model, but will take into account uncertainty and variability in the dose prediction.

Future work recommended includes further validation of the probabilistic modelling suite and the development of a standard procedure for conducting assessments with WATP and ADOP.

Main File

Some of the files on this site may be in a format that your computer can't read. However, you can download Readers and Viewers for the following document types below: