View Report Details
Investigation of the methods available for the practical utilisation of censored environmental measurement data
Project Code: R02008
National Radiological Protection Board
Daniels, W ;
Social Science Research Unit, Institute of Education, University of London
The Food Standards Agency (the Agency) supports a large surveillance monitoring programme measuring radioactivity concentrations in foods and other materials around nuclear sites. However, this data often provides an incomplete picture of the distribution of radioactivity in the environment due to the difficulty of detecting very low levels of radioactive contamination. It is inappropriate to spend a large amount of resources in cost and time seeking lower limits of detection (LoD) to circumvent this problem. In many cases this is practically impossible and in any event the radioactivity concentrations below the LoD used, are not, in and of themselves, a problem. However, the samples with radioactivity concentrations below the LoD only provide ‘censored’ information, in that their true value is hidden and all that is known is that the radioactivity concentrations are less than the LoD. This lack of knowledge can lead to distortions in important statistics such as the mean activity concentration in a food. In addition, due to this lack of knowledge, the Agency has conventionally applied a conservative and hence distorting assumption that all terrestrial measurements reported to be below the LoD should be treated as though they were at the LoD. This has adverse consequences, for example, it can greatly inflate estimates of individual and collective dose based on measured values.
Purpose of Study
The purpose of this study was to explore the options available to the Agency to improve the use made of measurements of radioactivity in food and the environment, as reported in the annual RIFE reports. In particular, the study reviewed the techniques available for establishing improved estimates of the summary statistics of the distribution of radioactivity concentration in the foods sampled by the Agency. A common assumption is that environmental concentrations will be lognormally distributed. This study therefore also examined the likely validity of this assumption, by reviewing the processes and mechanisms contributing to the distributional forms found in practice, with the aim of providing advice on the distributional form to assume when estimating relevant summary statistics from the data. In the light of the review findings the study aimed to provide advice on the most appropriate techniques to apply to give improved, robust and justifiable estimates of the distributional statistics of environmental radioactivity concentrations.
The review and study undertaken have revealed the likely mechanisms at work to produce the commonly observed environmental distributions of contamination data and demonstrated that the assumption of a particular distribution should not be made without background evidence or justifying argument. The study has shown that in many cases a lognormal distribution may be a characteristic consequence of the phenomena or a good approximation to the distribution generated by the physical processes under study.
The study has also discussed and explored a range of techniques that are available for handling censored environmental distributions. The approaches presented cover a range of options from the application of parametric techniques where all the sample values are assumed to be drawn from a known distribution, through robust methods that apply distributional assumptions only to the censored measurements, to fully non-parametric approaches where no distributional assumption is required. The techniques discussed also cover a range of complexity from simple substitution methods to advanced Bayesian methods. A number of techniques have been tested on RIFE data and the RMS errors of the predictions presented in the report. Some of the easy to use techniques of intermediate complexity were found to perform remarkably well even at extremely high censoring intensities and it is recommended that these simple to use approaches are considered for future use.
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: